Professional Documents
Culture Documents
Pulp and Paper2
Pulp and Paper2
Jennifer Peters
Department of Chemical Engineering
McCill University, Montreal
August, 1998
A thesis submitted to the Faculty of Graduate Studies and Research in partial fùlfillment
of the requirements of the degree of Master of Engineering.
The author has granted a non- L'auteur a accordé une licence non
exclusive licence aiiowing the exclusive permettant à la
National Library of Canada to Bibliothèque nationale du Canada de
reproduce, loan, distribute or sell reproduire, prêter, distribuer ou
copies of this thesis in microform, vendre des copies de cette thèse sous
paper or electronic formats. la forme de rnicrofiche/nlm, de
reproduction sur papier ou sur format
électronique.
Effluents from two rnills in Quebec were examined. Initially, effluents fiom a
chemithermomechanica1 pulp (CTMP) mill were used in this research. The objective was
to study the effect of six different waste streams mixed with total miIl effluent (0-10W)
on the reduction of COD and on the microbial population. The effluents fiom the ultra-
high yi eId pu1pi n ç process. wood washing process, acid plant, and paper machine were
found to have adverse effects on the biomass and its reduction of COD. The remainder of
this research focused on treating effluents fiom a Kraft mill and investigated the effect of
hydraulic residence time (HRT) and step inputs of 1%, 2.5% and 5% black liquor on the
reduction of COD and BOD5 and on the microbial population. Generally, the system
reduced the COD by 63-72% and the BODJ by 90-99%, regardless of HRT or black
liquor concentrations. Changes in HRT had no effect on the substrate uptake rate; it
ranged from 0.31 mgCOD/min*gVSSto 0.38 mgCOD/min.gVSS. However, the substrate
uptake rate did increase fiom 1.O mgCOD/min*gVSSto 1.6 mgCOD/min*gVSSwith an
i ncrease in biack liquor concentration. Funhermore, the SOUR increased fiom about 10
mg02/sVSS-h to 30 mg02/gVSS-h,indicating that the biack tiquor caused an increase in
microbial activity. The microbial population changed over time for different step inputs
of black liquoi- and the sludge quickly adapted to the new substrate loading. The Biolog
assay was an effective tool for indicating t hese changes.
Les déchets de l'industrie des pâtes et papiers sont couramment traités par des boues
activées afin de réduire le contenu organique et la toxicité des eaux usées. Dans la
plupart des applications industrielles, on doit traiter des effluents combinés provenant de
divers procédés. La composition de l'effluent combiné peut varier à cause de
déversements accidentels. L'objectif de cette étude est de déterminer les effets de ces
variations sur la communauté microbienne et sur la cinétique de la réduction des
demandes biologique et chimique d'oxygène @BO et DCO). Les résultats de cette étude
serviront à améliorer la compréhension des besoins de contrôle du procédé.
Des emuents provenant de deux usines du Québec ont été examinés. La recherche a
porté d'abord sur des effluents d'une usine de pâte chemi-thermomécanique. On a étudié
l'effet de six effluents différents mélangés à I'emuent total de l'usine (0% à 100%) sur la
population microbienne et la réduction de DCO. On a ensuite étudié le traitement des
effluents d'une usine Kraft. On a analysé l'effet du temps de résidence hydraulique
(TRH) et d'échelons de liqueur noire de 1%' 2.5% et 5% sur la réduction de DCO et
DBOz ainsi que sur la population microbienne. En général le système a réduit la DCO de
63?6 à 72% et la DBOs de 90% à 99% indépendamment du TRH et de la concentration de
liqueur noire. Les changemets de TRH n'ont eu aucun effet sur le taux d'utilisation du
substrat q u i a varié de 0.31 mgDCO/min*gSVSà 0.38 mgDCO/min*gSVS. Cependant le
taux d'utilisation du substrat a augmenté de 1.O mgDCO/min-gSVSa 1.6
mçDCO/min.gSVS suite à une augmentation de concentration de liqueur noire. De plus,
la liqueur noire a causé une augmentation d'activité microbienne tel qu'indiqué par le
taus d'utilisation spécifique d'oxygène qui a augmenté de 10 mgOz/h-gSVS à 30
mçOz/h-gSVS. La population microbienne a changé avec le temps pour différents
échelons de liqueur noire et les boues se sont rapidement adaptées à la nouvelle charge de
substrat. Le test Biolog s'est avéré un outil efficace pour indiquer ces changements.
ACKNOWLEDGMENTS
1 would like to acknowledge the following people for their heIp and support during the
course of my research:
Dr. Berk my research supervisor, for his advice and guidance and the Reaction
Engineering Research Group for their input and fnendship.
Sylvie Graffand Gordon Brodenck fiom Noranda Technology Centre for their time and
valuable input and for Noranda's direct financial involvement.
As well. I would like to thank the following students for their own research projects that
provided usefûl insight to my research:
Joëlle Milin
Menilick Grais
Car1 Mercadante
Sanjay Vadadoria
Final1y, 1 would like to thank my family and fiiends for always beinp there for me and
making my time at McGilI an enjoyable one. 1 would especially like to thank Jacint
D r a p for his support and understanding during the writing of this thesis.
TABLE OF CONTENTS
3 . 2 AVALYTICALPROCEDURES ...................................................................... 20
3 2 .I Tord Sqvetlded atrd C.O/ati/eSrrspetrded Sofids Coticet~tration(TSS and
I3S) 21
3.2.2 Cltemica/ Ox)lgett Dematrd (COD)............................................................. -21
3.2.3 Orgcrrric Carbon (TOC)....................................................................
TOICI/ 2 3
3 2 .4 5-Da). Biofogical OxygerI Demartd (;BOD_z) ............................................... 23
3.2.5 Spctcrjic Ovgett Uptake Rare (SOUR) ....................................................... 25
3.2.6 S/ttdge CWtme Itrdex (SI/Il........................................................................ 2 6
3.2.7 . .
pH Tetnperatttre Dissofved Ox)ge>r(DO) a»d Cotrdt~c~ivity . .................... 27
3.2.8 Riolog Microplde Techique atd Anabsis................................................ 2 7
3 . 3 EXPERIMENTAL PROCEDURES ................................................................. 33
3.3.1 Sbake F/as&Experimer 1fs............................................................................ 33
3.3.2 Reactor Erperime~rts................................................................................. 3 4
4 .
RESULTS AND DISCUSSION FOR EXPERIMENTS USING CTMP MILL
EFFLUENT .
.......................................................................................................... 3 8
4.1 PRELIMINARY EXPERIMENTS ................................................................... 38
4.1. 1 Slt~dgeImc~ihtmExperimmts .................................................................. 3 8
4.i 2 Ntttrierrt Expriments.................................................................................. 39
4.1.3 i2zriability Experime~rfs................
..... .................................................. - 4 1
4.I .4 Disctissiot~For Preliminary Experiments ....................... ..... ...............-I f
4.2 SHAKE FLASK EXPERIMENTS FOR COMBINATIONS OF INDIVIDUAL
STREAMS WITH TOTAL MILL EFFLUENT.......................................................... 42
4.2. I Paper Machine Efltent (PM) .................................................................. 4 3
4.2 . U/traHigh Yield Efllrent ((IHyI ................................................................ 43
4.2.3 Wood Washi~tg Efltie~rt0 .................................................................... 46
42 . 4 Scrre~rRoom Eflrtettt (SR).................... .- ................................................ 4 6
4.2.5 AcidP/arrtEflliec~t(AP) ....................................................................... 4 9
4.3 COMPARISON OF THE VARIOUS EFFLUENTS ......................................... 49
4.4 PRELIMINARY REACTOR EXPERIMENTS ..............................................5 1
4 - 4 1 CS TR tising Total Mi// Eflwrt withorit Recycle ......................................... 51
4.4.2 CSTR tisiqg Totd Mil/ Efltier u with Recjde ................................ . . ......5 2
Table 3.1 : Comparison of Experimental and Replicate variances for Selected Wells fiom
the Microplate .............................................................................................. 33
Table 4.1 : Vanous Sludge Inoculum Concentrations in Total Mill Effluent fiom the
CTMP Mill (time = 48 h) ............................................................................. 39
Table 4.2: Nutrient Experimental Results for CTMP Total Mill Effluent (time = 48 h) .
Varying the Nitrogen Concentration ............................................................ - 4 0
Table 4.3 : Nutrient Experimental ResuIts for CTMP Total Mill Effluent (time = 48 h) .
Varying the Phosphoms Concentration......................................................... 40
Table 4.3. Shake Flask Variability Results for CTMP Total Mill Effluent .10 Flasks ... 4 1
Table 4.5 : SV1 for the CSTR with Sludge Recycle using TME fiom the CTMP Mill
(HRT = 9.6 h) .............................................................................................. 53
Table 5.1 : Steady State Values and Ratios for an HRT of 12 h with Kraft Total Mill
Emuent ........................................................................................................ 58
Table 5.2. SV1 for an HRT of 12 h with Kraft Total Mill Effluent ................................59
Table 5.3 : Steady State Values and Ratios for an HRT of 20 h wit h Kraft Total Mill
Ernuent ........................................................................................................ 61
Table 5.3.SV1 for an HRT of 20 h with Kraft Total Mill Emuent ................................ 62
Table 5.5: Steady State Values and Ratios for an HRT of 8 h with Kraft Total Mill
Emuent ........................................................................................................ 65
Table 5.6. SV1 for an HRT of 8 h with Kraft Total Mill Emuent .................................. 65
Table 5.7: Steady State Values and Ratios for an HRT of 16 h using Kraft Total Mill
Emuent ........................................................................................................ 67
Table 5.8. SV1 for an HRT of 16 h with Kraft Total Mill Effluent ................................ 67
Table 5.9: Comparison of Ratios and Kinetic Parameters at Steady State for Various
HRT's .......................................................................................................... 69
Table 5.10: Steady State Values for a Step Input of 5% Black Liquor in Kraft Total Mill
Emuent (HRT = 16 h) ................................................................................. 75
Table 5.1 1 : SV1 for a Step Input of 5% Black Liquor in Kraft Total Mill Effluent (HRT =
16 h) ............................................................................................................ 76
Table 5.12: Comparison of Steady State Averages for a Black Liquor Step of 2.5% in
Kraft Total Mill Effluent (HRT = 16 h) ........................................................ 79
Table 5.13: SV1 for a Black Liquor Step of 2.5% in Kraft Total Mill Effluent (HRT =
16 h) ........................................................................................................... -79
Table 5.14: Cornparison of Steady State Averages for a Black Liquor Step of 1% in Kraft
Total Mill Effluent M T = 16 h) ................................................................. 83
Table 5.15. SVI for a Black Liquor Step of 1% in Kraft TME (HRT = 16 h) ........... 83
Table 5.16: Comparison of Ratios and Kinetic Parameters at Steady State for Various
Black Liquor Steps (HRT= 16 h. Sludge Age = 10 days) ............................. 86
LIST OF FIGURES
Figure 1 . 1 : Activated Sludge Plant Flow Diagram with a Single CSTR. ......................... 5
Figure 2.1 : Typical Growth Curve for a Single Bacterial Population Grown in a Batch
Reactor ......................
...
..- ....- -..... . . . ...... . .... .. ........ . .. .. . . ... . 12
Figure 3.1 : Typical Variation of DO with time as used for the Determination of SOUR26
Figure 3.2: Biolog GN MicroPlatem Panel ...... ..... ......... .............. ................................. 28
Figure 3 3 :Typical Biolog Microplate Image Taken using the Hi-Resolution Digital
Camera with an Overlay Showing the 5x5 Pixel Matrices at the Centre of Each
Well. ............................................... ......................-...-...----.-.--.--*-.....--.---..-...
30
Figure 3.4: Typical Scores Plot for the First Two Principal Components .......................32
Figure 3.5: Schematic of the Lab Top Continuous Reactor ..................................... . . .34
Figure 4.7 a and b: TSS and COD for the CSTR with Sludge Recycle Using TME corn
the CTMP Mill (HRT = 9.6 h) ...................................................................... 53
Figure 4.8: Scores for the First Two Principal Components for Microplates Inoculated
with SampIes Taken fiom Different Locations and at Different Times for
CTMP TME. (a = aerator, c = clarifier, s = sludge, numbers represent different
dates - t[l] is the scores vector for the first PC and t[2] is the second). ......... 54
Figure 5 . la, b and c: System Parameters for an HRT of 12 h using Kraft Total Mill
Effluent ....................................................................................................... -57
Figure 5.2 a, b, c and d: System Parameters for an HRT of 20 h with Kraft Total Mill
Effluent ........................................................................................................ 60
Figure 5.3: Scores for the First Two Principal Components for Microplates Inoculated
with Samples Taken tiom Different Locations at Steady and Non-Steady
States for the Kraft Mill Effluent at an HRT of 20 h (a = aerator, c = clarifier, s
= sludge, numbers represent dates -- t[l] is the first principal component and
t [2] is the second). ....................................................................................... .62
Figure 5.4 a, b, c and d: System Parameters for an HRT of 8 h with Kraft Total MiIl
Emuent ........................................................................................................ 64
Figure 5.5 a, b. c and d: System Parameters for an HRT of 16 h using Kraft Total Mill
Effluent ........................................................................................................66
Figure 5.6: Comparison of SOUR and SVI for Various HRT ........................................ 71
Figure 5.7: Scores for the First Two Principal Components (t[l] and t[2]) for Microplates
Inoculated with Samples Taken fiom the Aerator, Clarifier and Sludge at
Various Times for HRT's of 12 and 20 h using Kraft Mill Ettluent. ............. 72
Figure 5.8 a, b, c and d: System Parameters for a Black Liquor Step of 5% in Kraft Total
Mill Emuent (HRT = 16 h) .......................................................................... 74
Figure 5.9: Scores for the First Two Principal Components for Microplotes Taken at
Various Times Before and M e r a Black Liquor Step of 5% in Kraft TME
with an HRT of 16 h (t[l] is the first principal cornponent, t[2] is the second).
................................................................................................................... -76
Figure 5.10 a. b, c and d: System Parameters for a Black Liquor Step o f 2.5% in Kraft
Total Mill Efluent (KRT = 16 h) ...................... .......................................... 78
Figure 5.1 1 : Scores for the First Two Principal Components for Microplates Taken at
Various Times Before and M e r a Black Liquor Step of 2.5% in Kraft T M .
with an HRT of 16 h (t[l] is the first principal component, t[2] is the second).
Figure 5.12 a, b, c and d: System Parameters for a Black Liquor Step of 1% in Kraft
TME (HRT= 16 h) ...................................................................................... 82
Figure 5.13: Scores for the First Two Principal Components for Microplates Taken at
Various Times Before and After a Black Liquor Step of 1% in Kraft TME
with an HRT of 16 h (t[l] is the first principal component, t[2] is the second).
Figure 5.14: Effect of Black Liquor on the COD Uptake Rate for Kraft Mill Effluent.. .87
Figure 5.15: Scores for the First Two Principal Components (t[l ] and t[2]) for Images of
Microplates Inoculated with Samples fiom the Aerator Treating Kraft Mill
Effluent at Steady State before a Black Liquor Step (Replicates are also
shown). ....................................................................................................... -90
Figure 5.16: Scores for the First Two Principal Components (t[1] and t [ 2 ] )for Images of
Microplates Inoculated with Samples from the Aerator Treating Kraft Mill
Emuent at Steady State after a Black Liquor Step (Replicates are also shown).
....................................................................................................................91
NOMENCLATURE
sii
X biomass concentration (mass/volurne)
Y yieid coeficient (mass of biomasslmass of substrate)
subscripts:
a aerator
C dari fier
e effluent
f feed
r recycle
S substrate
W waste
X biomass
Greek letters:
a recycle ratio
/J specitic growth rate (tirne-')
Hni maximum speci fic growth rate (time")
0 sludge age (days)
l 1. INTRODUCTION
I
1.1 BACKGROUND
Biological treatment of wastewater is widely used in industry. One successfbl process
used extensively is the activated sludge process since it can effectively reduce the organic
content and toxicity of various wastewaters. The activated sludge itself is a complex
microbi al population that includes bacteria, protozoa, and many other types of organisms.
These microorganisms are mixed in suspension with the wastewater under aerobic
condit ions and metabolise the soluble organic compounds present. The bactena form
aggregates or flocs that can easily be separated from the treated wastewater in a settling
tank. Protozoa and other grwers feed on the edges of the flocs. Because of its complex
nature. the activated sludge process is dificult to control and extensive research is being
carried out to improve its control.
SeveraI factors, such as the reactor temperature, pH, hydraulic residence time (HRT), and
the nature and composition of the microbial population, rnay affect the performance of
any biological treatment system. Organisrns typically thrive at a narrow range of
temperature and pH. Changes in these parameters may lead to the growth of unwanted
organisms, such as filamentous bacteria, which cause bulking. Funhermore, most
activated sludge populations consist of a mixed culture, and the more diverse the culture
is the more adaptable the sludge is to different emuent streams. Therefore, the nature of
the sludge is an important consideration. Treatability of the effluent is firther
complicated by such factors as biodegradability, solubility of organic material, and the
presence of toxins. Toxins may include various bIeaching chernicals, trace metals, and
non-biodegradable compounds that could inhibit treatment.
Another factor that significantly affects the performance of the biological wastewater
treatment system is the composition of the wastewater itself. Most industrial applications
require the treatment of a combination of streams fiom various treatment processes. It is
impossible to keep the composition of the combined Stream constant over time since the
raw materials to the process may change, or variations such as shock loadings and spills
may occur. It is desirable to determine the effects of these variations on the microbial
community and ultimately on the kinetics of the reduction of biological oxygen demand
(BOD) and chemical oxygen demand (COD),thus leading to more accurate modeling of
the process.
The activated sludge process is commonly used in the treatment of pulp and paper wastes.
In the present research, effluents fiom two mills in Quebec were used. Mill A is a
chemithermomechanicd pulp (CTMP) mil l while Mi Il B uses the Kraft process. These
processes are briefly outlined below. The intention of this project was to investigate
several operating parameters that describe the performance of the activated sludge
process in order to determine the bioreaction kinetics. This will allow for more accurate
modeling of the activated sludge system and improve the understanding of control
requirernents for the process. A review of the treatability of the effluents by activated
shdge has been provided followed by the results and discussion for experiments done
using effluents from the two different rnills.
The Krafi miIl (Mill A) uses hardwood as its raw material. The main component is maple
(which constitutes about 60% of the total); however, birch and poplar are also used.
Hardwoods generally do not yield any extractives such as resin and fatty acids, which
contribute significantly to toxicity. However, the effluent may contain chlorinated
aromatics derived fiom the bleaching process. The Kraft mil1 uses a C&o.p.DED
bleaching scheme. Co refers to the use of CIOl (about 70%) and CI2 (g) in the first nage
of bleaching. Egp refers to extraction using O2and peroxide. DED refers to a second
bieaching stage using only CIO2,followed by extraction (without O2or peroxide)
followed by a final bleaching stage with CIO2. Chlorine derivatives from this bleaching
process maji be present in the miIl effluent. Also, plant upsets may result in a black
liquor spill that ends up in the final mil1 effluent and affects the treatment performance.
The effluent is characteristically dark in colour and tends to foam.
The CTMP mil1 (Mill B) uses soflwwd for its raw material. Sofiwoods typically contain
resin and fatty acids (RFA) as well as neutral wood extractives which give the final miil
emuent a high BOD and COD. Sodium sulphite fiirther increases COD since more RFA
are released. Sulphite is toxic to the biomass and high concentrations must be reduced
prior to bioiogical treatment. Different chernicals such a s peroxide ( H 2 0 2 ) and sodium
thiosulphate (NatS204) are also added to brighten the pulp depending on what type o f
paper is produced and the desired brightness. This affects the wastewater treatment
system further downstream. (Smook, 1992; Betz, 1980).
From the aerator the mised liquor (mixture of wastewater and biomass) is sent to the
clarifier at a rate of (Qr+ Q,) as s h o w in Figure 1.1. The flocculated biomass settles out
leavinç a final emuent vinually free o f solids. Some of the sludge is recycled to the
aerator at the rate. Q, t o provide continuous seeding and the remainder of the sludge is
wasted at a fixed flow rate, Q,v. Recycle keeps the concentration of the cells higher than
the ncrmal steady-state level and increases the conversion. It may also improve the
stability of the system by minimising the effects of process fluctuations (Curds, 1973).
Figure 1.I: Activated Sludge Plant Flow Diagram with a Single
CSTR.
subscript s = substrate
r, =@'
Models for the activated sludge process are typically based on the assumption that the
specific çrowh rates of the organisrns obey Monod kinetics and are related to their
specific substrate concentrations by the Monod Equation for endogenous metabolism,
(1.4)
where: p., = maximum specific growth rate (time")
K,= saturation constant, substrate concentration at one-half the
maximum growth rate (mass/volume)
S = limiting substrate concentration (mass/volume)
kd = decay rate (tirne-')
As an organism grows, it is assumed that the rate at which the organism uses substrate is
described by the following relation:
The sludge age can be defined as the average length of time that a microorganism or
sludge particle spends in the aeration tank. If the proponion of microbial cells is assumed
constant than sludge age may be considered equivalent to çolids retention time (SRT).
SRT is usually controlled through sludge wasting by the following equation:
KX
SRT =
O,TSS, - Q,.X,
Effluents from two different pulping mills were examined in this study. Initially, the
intention was to use only emuent fiom a CTMP mil1 (Mill A) for this research. However,
since emuent from Mill A became unavailable due to changes in the mill's status within
the Company, the scope had to be altered to include the study of emuent from a second
mil1 (Mill B) employing the Kraft process. The original plan of the research was to
characterise the CTMP mil1 effluent using shake flask experiments and to ascertain which
of its constituents had the strongest effect on the overall reduction of COD. Once the
waste streams with the most significant effect were identified, testing could be limited to
these streams and experiments in a continuous reactor could be performed.
However. afier only two preliminary experirnents were performed in the continuous
reactor. effluent was n o longer available as stated above. Thus, experiments were
undertaken with emuent samples fiom Mill B employing the Kraft process. Several
different hydraulic residence times were investigated using a continuous reactor with
sludge recycle to determine the effect of HRT on the reduction of COD and BODs. Once
these were completed, a series of step experiments adding black liquor at various
concentrations were performed to determine the effect of black liquor on the microbial
population and its ability to reduce the COD and BODs.
1.2.1 Statement of Objectives
Mill A
1. To identify which o f the emuent streams have the greatest effect on the biomass and
its abiiity to reduce the COD.
2. TOcharacterise the effluent and sludge using shake flask expenments and the
following analytical techniques: COD, TOC, BODJ, TSS,VSS, SVI,and phenotypic
fingerprinting using the Biolog technique
Mill B
3 . To determine the kinetics of the reduction of COD and BOD5 by the activated sludge.
4. To study the effect of black liquor on the sludge population and the reduction of COD
and BODz to improve the understanding o f what occurs dunng a system upset.
l 2. LITERATURE REVIEW
I
2.1 MICROBIOLOGY OF ACTWATED SLUDGE
Activated sludge consists of a diverse collection of microorganisrns. Organisms require
energy and carbon sources for cell production and proper fûnctionality. The carbon may
be derived fiom organic matter or carbon dioxide. Cells using CO2 are classified as
autotrophi c while cells that utilise organic carbon are heterotrophic. Microorganisms can
be metabolically classified by the energy source they require. Photosynthetic autotrophs
use light as their energy source while chemosynthetic autotrophs derive their energy from
inorganic oxidation-reduction reactions. Heterotrophs derive their energy fiom organic
osidation-reduction reactions. AI1 microbes can be ftnher classified by their oxygen
requirements. Aerobic species need oxygen for growth, anaerobic species need an
environment fiee from oxygen, and facultative species can thrive with or without oxygen
present (Metcalf and Eddy, 1979).
Protozoa are single-celled, motile organisms that include amoebas, flagellates, fiee-
swimming and stalked ciliates. These feed on the bacteria and help maintain a balance
wit hin the population. Most protozoa are aerobic heterotrophs. Another group
sometimes present in healthy sludge are rotifers. Rotifers are aerobic, heterotrophic,
multicelluiar animals that consume dispersed and flocculated bactena as well as some
organic matter. They help improve the efficiency of the treatment process (Metcalf and
Eddy, 1979). Aerobic organisms grow best in non-limit ing dissolved oxygen
concentrations greater than 6.0 mgA, but a dissolved oxygen concentration between 1-2
mg/l is suficient for most active aerobic heterotrophic microbial activity (Gray, 1990).
It is best to operate activated sludge reactors in the endogenous growth phase. Dunng
this phase, the total mass of rnicroorganisms begins to decrease as cells use up stored
reserves and begin to die. This leads to good BOD removal because the microbes are
oxidisinç a11 available substrate. Also, floc-forming bacteria are only flocculant in a
stationary growth phase such as the endogenous (or declining) phase (Klopping et al,
1995; Gray, I W O ) . Furtherrnore, the log growth phase promotes poor settling conditions
for the sludge, which leads to a higher concentration of solids in the effluent as well as
inefficient BOD removal (Barret al, 1996).
Timc 01)
Using sequencing batch reactors, Franta et al (1994) studied factors afEecting the
composition and concentration of residual organics in the emuent of biological
wastewater treatment plants. They tested mechanical pulp mil1 wastewater and a
synthetic wastewater to estirnate the fraction of residual organics in the effluent
originating fiom activated sludge metabolism. They determined that only 10% of effluent
(residual) COD had microbial origin. The rest of the residual organics could be assumed
to corne from higher molecular weight compounds derived from lignins. The authors
concluded that the best COD removal and sludge settling for papermill wastewater were
achieved with the highest sludge age (20 d) and longest react period (22 h).
Franta and Wilderer (1 997) studied the reduction of residual organics using sequencing
batch reactors by modifying the sludge age, react period (batch phase) and fiIl period. In
terrns of biodegradability, they generally achieved a reduction in COD between 8 5 9 3 %
and a 98% reduction of BODs, while the SV1 (sludge volume index) ranged from 90-170
m l Low molecular weight compounds such as alkenes, terpenes, diterpenes, resin
acids. aromatics and aliphatics were significantly reduced in the effluent indicating that
they are readily biodegradable. Higher molecular weight fiactions, such as
2-methosyphenol derivatives, required high sludge açes to achieve significant
degradation ratios. They found that the system achieved the minimum residual COD at
shdge ages of 15 and 25 days. Overall, they achieved the best residual COD removal
with a sludge age of 15-25 days, a react period of 12 h and a fiIl phase of 0.5 h.
compiling al1 availabie data in order to get clues on which substances may be present in
\
the emuent. Then techniques including simple chernical analyses such BODs, COD,
VSS, and the determination of ammonia and ad hoc group specific parameters should be
applied. Finally. for aerobic stabilisation he suggested that investigators should assess the
degradable and persistent toxicity as well as the overall 'ready' biodegradability.
A technique recently developed by Biolog Inc. in Hayward, California has been reponed
by several research groups (Fulthorpe et al, 1992; Fulthorpe and Allen, 1994) to be
effective in the characterisation of bactenal communities in the activated sludge when
coupled with statistical analysis. The Biolog technique uses micropIates comprised of 95
wells each containing different sole-carbon sources and tetrazolium violet, a redox dye
that tums purple upon respiration. An additional well containing no nutrients acts as a
control well. The pattern of purple weIls may be analysed to determine which species
was inoculated on the plate (a more detailed description of the Biolog technique is given
in Materials and Methods, section 3.2.8).
Garland and Mills (1991) used the Biolog technique to study the rate of colour
development of various microbial communities. They applied principal component
analysis (PCA) to determine intrinsic relationships. PCA is an ordination technique that
essentially rotates a swam of data about its centre and projects the data ont0 new axes or
principal components. This analysis extracts the greatest variance in the fira two
principal components and these may be plotted in two dimensions to reflect any intnnsic
patterns within the data (see Appendix B for a complete description on PCA).
Pnor to the use of PCA, Garland and Mills (199 1) rnanipulated the data. First, they
determined the raw difference values for each response well within a plate by subtracting
the colour response of that well from the colour response of the control well (C-R). Then.
the data was transformed by dividing the raw difference values by the average well colour
development (AWCD) : (C-R)/([Z(C-R)]/9S).
This step is necessary to eliminate
discrepancies among plates due to variations in inoculurn density and the rate of colour
formation (Garland, 1996). They incubated whole communities in Biolog GN plates (a
type of plate thar contains various substrates typically util ised by the gam-negative
bacteria inherent in the activated sludge) and read the colour response with a digitising
video scanner. They determined that inoculum cell density affected the colour
development rates and suggested that colour change in the Biolog community level assay
is caused by growth of bacteria, rather than respiration.
Schneider et al (1998) determined that fingerprinting can indicate a shifl in the microbial
pattern during system upsets such as a toxic breakthrough and was, in fact, more sensitive
than conventional plate counts and microscopy. Thus, the authors confirmed that
phenotypi c fingerprint ing was sensitive enough to detect changes that affect the system
performance. They also used PCA. The authors measured the colour response of each
well and obtained the colour development for a particular well by subtracting the colcur
response of that well at inoculation fiom its colour response afler a 24 h incubation (C-R).
This data was considered the raw difference data. The overall colour development of the
plate u a s espressed as AWCD : [Z(C-R)]/95].
The data was transformed by dividing the
raw difference value for each well by the AWCD of the plate [(C-R)/AWCD]
to express
the pattern of sole-carbon source utilisation of that plate. The authors then performed
PCA on the transformed data to determine the relationships among different samples
(plates) and found distinct differences between samples taken d u h g a toxic upset and
sarnples taken during normal operation.
3. MATERIALS AND METHODS 1
Pulp mill efïluents were received approximately once per week in shipments of about 16-
32 litres. These shipments were transponed in coolers with ice packs to keep the
effluents cool and minimise changes in emuent properties. Sludge was shipped in the
same manner (about 2-4 litres per week or as needed). Sludge and effluent were
refrigerated immediately upon receipt and stored at 24°C until needed (sludge was used
as soon as possible to ensure that it was always fresh and that the ceflswere still viable).
P i o r to use, the sludge was allowed to settle and then the supernatant was decanted to
concentrate the sludge. Total mil1 effluent (TME), which is the combination of al1 the
waste streams at the mill, was taken afier prirnary settling that removes most of the large
fibres and particles while sludge was taken fiom the retum activated sludge line.
Constituent emuents were taken fiom their respective points prior to joining the total mil1
effiuent.
Chemical osygen demand (COD) is the oxygen equivalent of the organic and inorganic
matter in a sample thet may be oxidised by a strong chernical oxidant. The closed reflux,
titrimetric method from Standard Methods (1986) was used for this procedure. It
involves refluxing the samples in a strong acid environment at a high temperature with a
known excess of potassium dichromate. The remaining dichromate is measured by
titration with ferrous ammonium sulphate (FAS). Samples were filtered and stored at
2-4°C for not more than two weeks before analysis.
It was necessary to prepare several reagents a day or two before beginning the assay to
allow time for the solutions to dissolve and cool. Standard potassium dichromate
digestion solution was prepared to a concentration of 0.0167 M by adding 4.913 g of
previously dried K2Cr20, and 167 ml of concentrated H2S04to about 500 ml of distilled
water, cooling and bringing the mixture to 1000 ml. Sulphuric acid reagent was prepared
by adding silver sulphate (Ag2S04)to concentrated H2SO4 at the rate of 5.5 g AgzS0.d kg
H2S04 (silver sulphate is added as catalyst to improve the oxidation of straight-chain
aliphatic compounds). Standard FAS titrant was prepared to a concentration of about
0.0 1 M by dissolving 3.92 g of iron (II) ammonium sulphate hexahydrate
(Fe(NH&(SO&-6H20) in distilled water and adding 20 ml of concentrated H2SO4,
cooling and diluting to 1 0 0 0 ml.
Standard 10 ml ampules and caps were pre-washed with 20% H2S04 to prevent
contamination. The ampules were prepared by combining 2.5 ml of sample, 1.5 ml of
digestion solution and 3.5 ml of sulphuric acid reagent. Appropriate caution was taken
when working with the strong oxidising and acid reagents. It was necessary to dilute the
sampIes (the filtrate collected during the tests for TSS and VSS) so that the resulting
COD would be within the range of 100-300 m g . This range was acceptable for this test
and it avoided the necessity of using very small volumes of the titrant. Four blank
ampules containing distilled water instead of sarnple were also prepared. Two of these
were digested to provide a blank reading and two were set aside to determine the molarity
of the FAS. The ampules were digested on a heating block preheated to 150°C for 2 h
then cooled to room temperature prior to titration. 1-2 drops of ferroin indicator were
added and the samples were titrated with FAS whiie stirring. The observed end point was
a change from blue-green to reddish brown. The COD was determined by the foltowing
equation:
Standard Methods (1986) Iists some possible interferences in the COD test. Silver
sulphate, added as catalyst, reacts with halides to produce precipitates that are orily
panially oxidised. Mercuric sulphate may be added to overcome complications due to the
presence of halides, but in this case it was assumed that the presence of halides in the
samples was insignificant. Also, interference due to nitrite (NOz') was insignificant.
Standard Methods ( 1986) suggests doing several controls to ensure the quality of the
dilution water and seed. One such control is the seed control. A seed control was
performed by plottinç the DO depletion venus volume of seed. This plot should produce
a straight line of which the slope is DO depletion per ml of seed and the y-intercept is less
than O. 1 mg/]. To d o this the DO depletion over five days was measured for six dilutions
of 3. 5, 10, 15. 20 and 25 ml o f seed. Two blank bottles containing only dilution water
were also incubated and the DO depletion was measured. It was usually less than 0.2
m g l , ensurin- h i ~ quality
h dilution water. Another check involved incubating two
bottles containing 3 ml o f seed and 6 ml of standard glucose-glutamic acid solution.
According to Standard Methods (1986). the BOD5 should be about 204 mg/l.
AI1 samples from the aerator were filtered as descnbed by the TSS method (section 3.2.1)
and stored in a freezer until the BODs assay could b e performed. Then the samples were
thawed and diluted by estimating the BOD5 and adding the required volume o f sample to
a standard 300 ml BOD bottle. 3 ml o f seed were added and the bottle was filled with
dilution water. The initial DO was rneasured by inserting the DO probe and stimng
mechanism into the BOD boale and waiting until the reading stabilised before recording
the value. Once the initial DO was detennined, the bottles were topped off wit h more
dilution water, stoppered and covered with parafilm to create a water seal and prevent
evaporation. The samples were incubated in darkness at room temperature for five days
before the final DO was measured as before. The BODs was calculated by the following
equation:
(3-3)
where: BOD5 is expressed in mg/]
Dl = DO of diluted sample immediately afier preparation (mg/l)
Dl = DO of diluted sample after 5 d incubation at 20°C (mg/l)
P = decimal volumetric fiaction of sample used
Bi = DO of seed controt before incubation (mg/l)
Bz = DO of seed control after incubation (mg/])
f = ratio of seed in diluted sample to seed in seed control
A. Microplate Preparation
The Biolog technology relies on the irreversibie reduction of a tetrazolium dye as an
indicator of carbon-source utilisation. There are several types of Biolog microplates that
may be used depend ing on the type of organism to be tested. Since activated sludge
consists mainiy of gram-negative species (Gray, 1990), GN microplates that are specific
to sram-negative species were used. A GN microplate panel is comprised of 95 substrate
containing wells and a control well without a carbon source. A graphical representation
of a GN microplate is given in Figure 3.2. Each well is also supplied with dye and
nutrients in a dried-film form, which is reconstituted upon inoculation of the cell
suspension. If the cells utilise the substrate, the tetrazolium redox dye tums purple.
Patterns of metabolic response for a community are produced, and these patterns can be
read usin2 imaging software and then analysed statistically to determine the community
response to the treatment. This is described below.
The microplates were stored a? 24°C and warmed to room temperature pnor to use.
Fresh samples of mixed liquor were collected from either shake flasks or the reactor and
deflocculating agents, sodium pyrophosphate and Tween 80, were each added to the
sample at a concentration of 0.01% v/v. The samples were homogenised by vigorous
shaking and any remaining solids were removed by centrifugation at 5000 rpm for five
minutes. The recovered suspensions were washed three times with 0.1 M phosphate
buffer by centrifugation at 10 000 rpm for ten minutes and then resuspended in about 2 ml
of 0.85% saline. This method was adapted from that described by Victorio et al (1 996).
The turbidity range was established by using turbidity standards provided by Biolog and
an uninoculated saline tube to set the 100% transmittance. The turbidimeter was re-set to
100% transmittance with the same uninoculated saline tube (this should be done for each
tube) and the cell suspension was slowly added until the inoculum density fell within the
acceptable turbidity range. The approximate ceIl density was adjusted to 3x10' cells/ml
and this gave a transmittance level of about 53% to 58%. Al1 microwells were filled with
precisely 150 pl of suspension, taking caution not to splash or carry over any chemicals.
The plates were incubated in a closed box at room temperature for 48 h. M e r 48 h,
pictures of the microplates were taken using either a low-resolution Kodac DC40 'point-
and-shoot' digital camera or, in later experiments, a Sony Hi-Resolution CCD-IRIS
monochrome dioital camera interfaced to a computer.
B. Image Processing
Initial images were obtained using a hand held, low-resolution camera. An attempt was
made to hold the camera at the same height each time; however, the distance between the
camera lem and the microplate may have varied from one imase to the next. Also, the
lighting was not as consistent since a fibre optic light source was not yet available. The
pictures were al1 taken in the same room with the blinds drawn in an attempt to minimise
inconsistencies in the lighting; however, there were some problems with glare and
reflection. In order to avoid this lack of precision, a more standardised procedure using
the Hi-Resolution camera was developed. The camera was mounted on a stand at a
constant distance from the microplate. The microplate itself was placed in a box with
white walls and a lid to prevent outside light fiom entering and creating glare and
reflection. A hole was cut into the lid to accommodate the camera lens and two fibre
optic light sources were directed through the hole and aimed at the walls inside the box to
produce consistent lighting.
The data were analysed with the aid of the Visilog 5.1 image analysis software. A typical
image is represented in Figure 3.3. Colour intensity in each well was determined using a
macro that computed the average pixel value of a 5x5 matnx of pixels taken fiom the
centre of each well (shown as small squares in the figure). These values were pretreated
following the procedure outlined in a paper by Garland and Mills (1991). This involved
computing the raw difference data by subtracting the colour intensity for the response
well fiom the colour intensity for the control well (C-R, where C is A l in Figure 3.2 and
R is each response well). Then each value was divided by the average weil coIour
development (AWCD) to normalise the data: AWCD = L(C-R)/95. This data, refened to
as the transformed data, was then statistically analysed using principal component
anal ysis (PCA).
Figure 3.3: Typical Biolog Microplate Image Taken using the Hi-
Resolution Digital Camera with an Overiay Showing the 5x5 Pixel
Matrices at the Centre of Each Well.
C. Statistical Analysis
PCA is a multivariate method that essentially rotates a swarm of data points about their
centroid to reveal any intrinsic patterns. SIMCA-Pstatistical software was used to
develop the PCA models. It follows the nonlinear iterative partial least squares
(NTPALS) method of calculating the principal components. The method consists of
rewriting a data matrix as a sum of linearly independent matrices. These can in turn be
expressed as products of two vectors, a score (column) vector, r and a loading (row)
vector. Once the vecton, t and are found they can be ploned two dimensionally to
reveal any intrinsic patterns present in the data. A plot of the score vectors, 2, and ta is
usuaily al1 that is required to demonstrate differences among plates. Scores are vectors
that contain the greatest variation that can be explained among the data. Score plots
reveal trends, groups, and outliers (observations that lie beyond the region within the
Hot elling t2 confidence ellipse). Loading vectors indicate which of the variables are
important in the approximation of the data matrix. Loading vectors may also be ploned;
however, these figures are difficutt to read and it is more effective to simply list the
important variables.
.A typical plot of the scores of the first two principal components (PC's) is given in Figure
3.4. This figure should be read in the following manner. The x-axis labelled t[l] is the
scores vector for the first PC while the y-axis labelled t[2] is the scores vector for the
second PC. The points in the figure are labelled according to experiment (by capital
letcers) and time the sample for the microplate was taken. In this particular figure, the
letter 'G' happens to represent a 5% black liquor step experiment. The times are with
respect to the addition of black liquor. Therefore, 'Oh bef signifies a microplate done at
steady state just before the addition of black liquor, while 'Oh afl' signifies a microplate
done immediately after the black liquor step was initiated. The points denoted by '14 h'
and ' 6 5 h' represent plates done at 14 and 65 h afier the step. These times may vary
slight 1y for different experiments. Finally, the point denoted by 'SS new' represents a
plate done when t h e system reached a new steady state in the presence of black liquor. In
this particular figure, the points are al1 slightly apart fiom one another indicating that the
pIates were chançing over tirne. It shouId be noted that each figure represents the scores
for each set of plates in a particular mode1 and separate models were developed for
individual experiments as well as for certain plates from several experiments in order to
compare them.
'G Oh aft
N O
GOhbcf ,. G ( j g h
GlSh
Figure 3.4: Typical Scores Plot for the First Two Principal
Components
For the hydraulic residence time experiments, slightly di fferent notation was used to
represent points for individual plates. The experiments are stilI distinguished by using
different capital letters; however, in these experiments the sarnples for the microplates
were taken on di fferent days and fiom different locations. A typical point wilI be denoted
by say 'B6s1,where the 'B' represents that particular experiment, the number '6'
represents the date the plate was done, and the letter 's' represents the location from
which the sample was taken, in this case, the refiigerated slud_ee. The letters 'a' and 'c'
represent samples taken from the aerator and clarifier respectively.
For the anal ysis. the data rnatrix was set up by placing the intensity values for each well
into a single row. Therefore, rows represent individual plates while columns represent
C-
the colour intensity in each microwell. In this way. differences among the plates could be
determined. However, PCA cannot test for specific differences. More details on the
setup of the data matrix and the pretreatment of the data, as well as a mathematical
description of the PCA model, are given in Appendix B.
In order to determine the variation of the PCA scores, a variability study was performed
by two different researchers who each prepared several microplates inoculated with the
same sample of sludge. The variance of the readings for each of the 95 wells in the
variability studies was very small indicating that the Biolog assay is very reproducible.
The variance for selected wells in one of the variability studies is much smaller than the
variance for the same wells in a senes of plates done during an expenment (see Table 3.1)
and the ratio of experimental variances to replicate variances was p a t e r than two for
97% of t h e microwells.
Once the nutrients and sludge were mixed with the wastewater (referred to as a batch),
about 100 ml of each batch were poured into one of several flasks and placed on the
rotary shaker; duplicate flasks were prepared for each batch. Initial samples of about
10 ml each were filtered in duplicate immediately upon mixing the batches for
determination of actual initial TSS,COD and TOC. TSS,COD and TOC were
det ermined at prescribed time intervals.
r s q clé
I
mised liquor
liml ireaiéd
cmucnt
Clarifia
I air 1
sludge
>
As in the flask experiments, it was necessary to determine the TSS of the sludge in order
to determine the amounts of sludge and effluent to combine to get the desired sludge
inoculum concentration. A batch of emuent was prepared, as described for the shake
flask experirnents, bg first adding nutnents to the wastewater and then adding Antifoam
A" (Dow Coming, Inc.) at the rate of 100-200pl per 2000 ml of effluent (more antifoam
had to be added when black Iiquor was to be used). The effluent was poured into the
reactor and the required amount of siudge was added to bring the final working volume to
1.4 litres. Agitation and air were applied and a few minutes were allowed to pass for
cornpiete misin3 before a sample was taken fiom the aerator and filtered in duplicate as a
time zero measurement. The filtrate was collected for measurement of COD and TOC
and the filter paper holding the solids was used to determine TSS and VSS.
b) Continuous Phase
The continuous phase started just as the sludge entered the endogenous phase. The pump
speed was adjusted until the desired flow rate was achieved. The continuous phase began
afier approximately 16 h of acclimation (batch phase). Afier about 14-16 h on the shaker,
the flasks were removed and poured into the clarifier and enough time was allowed for
the sludge to settle (about 1 h). Another sampte was taken just afler the continuous phase
began.
Sludge age was adjusted to the desired value by wasting fiom the clarifier once per day
based on the previous day's TSS in the aerator and clarifier with the following equation:
Vd TSS,
WSF =
O * TSS,
The Biolog microplate analysis was performed when the reactor reached steady state and
whenever a significant change in the population may have occurred (Le. when adding
black liquor).
In this chapter, the expenmental results and discussion for the treatment of effluents from
a CTMP mil1 in Quebec will be presented. Initially, preliminary experiments were
executed using shake flasks with total mil1 emuent. The results and discussion for these
experiments are ~ i v e nin section 4.1. Then shake flask experiments with constituent
emuents from the CTMP mi11 were performed. These results are given in section 4.2 and
the experiments are compared in section 4.3. Finally, preliminary expenments were
performed using a CSTR. These results are provided in section 4.4 and the discussion
follows in section 4.5. Further continuous reactor experiments were not camed out
because the effiuent was no longer available.
Initially. two experiments were performed in shake flasks using different nominal sludge
inoculum concentrations. In these experiments, total mil1 emuent (TME) from a CTMP
miII was examined to test the effect of various sludge concentrations on the reduction of
chemical oxygen demand (COD).The C0D:N:P was lOO:).O: 1.0, as described in
literature (Franta et al, 1994; Klopping et al 1995; Gray, 1990; Grau. 1991). The flasks
were placed on a rotary shaker for 48 h at 2 1-24OC with an approximate speed of
1 8 5 r p m The results for different sludge concentrations are shown in Table 4.1.
Relatively high percent reduction in COD was achieved, regardless of the sludge
inoculum concentration. The COD was generally reduced from initial concentrations of
750-1000 mg1 to final concentrations of 250-300mg/l. Based on this observation, a
nominal sludge inoculum concentration of 2000 mg/l was chosen for subsequent
experiments to reflect industrial conditions. The growth of the sludge was less than 35%.
A higher growth rate was not expected since initial inoculum of sludge was high and the
effluent, limited by the substrate concentration, could not suppon funher growth. The
effect of time on COD reduction was also examined and it was noted that most of the
reduction in COD occurred within the first 24 h, &er which there was no decrease.
Similarly, the sludge concentration did not change significantly between 24 and 48 h.
In order to determine the appropriate amount of nutrients to add to the wastewater fiom
the CTMP mill, an experiment was executed where nutrients were added on the basis of
effluent COD by the ratio C0D:N:P. In the first part of the experiment, the nitrogen
concentration was varied while the phosphoms concentration was held constant and in the
second part, the nitrogen was held constant while the phosphoms concentration was
varied. The experiments were performed using shake flasks inoculated with 500 and
2000 mg1 of sludge using TME from the CTMP miIl and incubated on a rotary shaker at
room temperature for 48 h. The objective of this test was to determine the best C0D:N:P.
Results are given in Table 4.2 and Table 4.3.
The best C 0 D : N ratio appeared to be 100:3 for both inoculum concentrations because it
achieved the greatea percent growth and reduction in COD. The best C0D:P ratio with
respect to the reduction in COD appeared to be only 100:0.2 for an inoculum
concentration of 500 mg/]while it was 1002 for an inoculum concentration of 2000 m d .
This ratio was higher than expected which suggests that for the system in q u e s t i o ~the
industrial ratio is too low with respect to the amount of phosphorus. Also, greater growth
occurs with a ratio of 100:2 and an inoculum concentration of 2000 mg/l. In general,
since most subsequent experiments used the inoculum concentration of 2000 mgA, the
best C0D:N:P ratio for CTMP mil1 effluent was deemed to be 100:3:2.
Table 4.4: Shake Flask Variability Results for CTMP Total Mill
COD
Reduction
1526 7-+
1491
523
506
489
489
45 4
489
506
47 1
506
540
The nutrient experiments indicated that the optimal ratio should be C0D:N:P of 100:3:2.
Nutnent addition displayed a dependence upon the initial biomass concentration in the
systern. More sludge required more phosphorus and nitrogen.
The individual streams were combined with TME in concentrations of 0, 5, 10, 20, 40,
60, 80, and 100% by volume and placed in shake flasks on a rotary shaker at about 180-
200 rpm for 48 h at room temperature. Concentrations greater than 60% for each
individual stream are seldom or never realised on an industnal scale; however, these
concent rat ions were examined in the experiments to investigate the effect of each
undiluted stream. In each of the following experiments, a high nominal inoculum
concentration of 2000 mg/l was used since the growth rate of the organisms was not of
primary interest in these experiments. The C0D:N:P was 100:3.0:2.0as determined in
the nutrient experiments. The results presented in the subsequent figures include the
COD, percent reduction in COD, TSS, percent growth in TSS and the COD utilised by
TSS at the experimental time of 24 or 48 h.
4.2.1 Paper Machine EiTlucnt (PM)
The paper machine effluent (PM) had a COD of 690 rng/l, which was lower than the
TME at 2300 mg/l; therefore, when the TME was diluted with the PM, the COD
decreased as s h o w in Figure 4. la. The percent reduction in COD varied between 60%
and 80% for both 24 and 48 h as shown in Figure 4. l b with no significant improvernent in
the reduction of COD over time. Figure 4. lc and Figure 4. Id indicate that as the
percentage of PM increased the totd suspended solids (TSS)decreased fiom 2200 mg/l to
1900 mg/l and percent growth decreased fiom 22% to about 8%. This suggests that the
paper machine effluent had some toxic effects on the sludge. From Figure 4. le, it is
evident that even low concentrations of PM had a detrimental effect on the specific
reduction of COD (ACODITSS). The specific reduction of COD dropped fiom 0.74 to
O 30 ~ v i t ha concentration of only 5% PM. This may be due to toxic effects of the PM on
the biomass.
The ultra high yield effluent (UHY) had a higher COD (about 2850 mg/l) than the TME
(about 1630 rngl);therefore, when the TJHY was added to the TME, the COD increased
as s h o w in Fisure 4.2a. As the concentration of UHY increased the percent reduction in
COD, TSS and percent growth al1 generally decreased as illustrated in Figure 4.2b, c
and d . Most notably there was no growth within the first 24 hours when UHY was
greater t hen 20%. After 48 h, however, onIy concentrations of 80% and 100% prevented
C-
erowth. This suggests that the ultra high yield effluent may have had initial adverse
C
effects on the sludge, but that the sludge adapted over time. The results indicate that a
longer reaction time of 48 h allowed the system to remove slightty more COD. Also, the
specific reduction in COD (ACODITSS)decreased from about 0.65 to 0.30 at
concentrations of UHY greater than 40% as shown in Figure 4.2e. This may fùrther
suggest that the UHY adversely affected the biornass.
a COD 1 1 b Y. RcRcbction in COD
i
0 - j i
O 50 100 1
% PMinTME ./. PMinTME
p
V
zoo0 - m a
I
4 :
I
cn 1900 - O o
O
O 9 i
O
L o o - O . .
% PMinTME
e. COD Uilissrfioci
O 50
Lrgend
The COD for the TME and the SR were both in the range of 1 100-1200 mg/]. Therefore,
as can be seen in Figure 4.4a. the COD did not Vary significantly with increasing amounts
of screen room emuent. Likewise, the reductions for the mixtures were in the same range
of 70 to 80% as s h o w in Figure 4.4b, and the TSS were generally in the same range of
2200 to 2350 mg/l as shown in Figure 4 . 4 ~ .The percent growth in each mixture was in
the same range of 1 5 to 20% at concent rations of 5% SR or greater. The specific
reduction in COD (ACOD/TSS) is illustrated in Figure 4.4e and the data remain fairly
constant wit h increasing concentrations of SR. Therefore, the SR had Iittle effect on the
reduction of COD or on the biomass. In al1 the figures, the 48 h trends were generally
sliehtly less in magnitude than the 24 h trends.
a COD b ./. Rcbctiorr in COD
Lrgend
O 1=zero
O i j
I
O 50 100 5O
% SRinTME % SRinTME
O
O 50 100,
% SRinTME
The acid plant effluent (AP) had a very low COD of approximately 50-60 mgIl thus the
COD decreased with increasing amounts of AP as s h o w in Figure 4.5a. The 24 and 48 h
COD are in the same range, regardless of AP concentration. This can be explained by
Figure 4.5b where the reduction in COD decreases fiom 75% to zero for higher
concentrations of AP. This is due to the fact that at high concentrations of AP there was
~ can be xen that the
very little organic matter for the sludge to consume. Figure 4 . 5 it
TS S decreased from 2 150- 1850 mg/i with increasing amounts of AP. This could be due
to tosicity of the AP on the cells, but rnay also be explained by the very low COD of the
AP. This effect is further evident in Figure 4.5d where the percent growth drops fiom
20% to zero with increasing amounts of AP. Figure 4.5e shows that the specific
reduction in COD (ACODTTSS) decreased fiom 0.42 to zero with increasing
concentrations of Ai? These observations indicate that the acid plant effluent provided
very little organic matter for sludge consumption and may also have been toxic to the
sludge.
In most cases, higher concentrations of the effluent (usually greater than 40%) inhibited
-growth (Figures 4.1, 4.3 and 4% and d). With increasing concentrations of the ultra high
yield effluent, however, the growth rate simply slowed d o m (Figure 4 . 2 and
~ d). In the
mill, the UHY accounts for almost half the TME; thus the sludge was Iikely acclimated to
it at lower concentrations. In the case of the wood washing and acid plant emuents, the
COD concentrations were quite low (Figure 4.3a and Figure 4.5a). Therefore, the sludge
may simply have run out of substrate rather than experiencing adverse conditions. The
screen room effluent had no effect on the biomass (Figure 4 . 4 and
~ d).
a COD b ./. Rebctioa in COD !
% AP in TME
% AP in TME
Figure 4.6 a and b: TSS and COD Cor the CSTR without Sludge
Recycle Using TME from the CTMP Mill (HRT = 28 h)
1800 -
1600
I
-
A
1400 4
400
zoo +
Figure 4.7 a and b: TSS and COD for the CSTR with Sludge Recycle
Using TME from the CTMP Mill (HRT = 9.6 h)
Table 4.5: SVI for the CSTR with Sludge Recycle using TME from
the CT3IP Mill (HRT = 9.6 h)
Time (h)
During the experiment the temperature, pH, DO and SOUR were monitored. The
temperature remained relatively constant at 24"C, the p H was in the range of 5.5-7.5 and
the DO was about 6 to 8 m d l . The SV1 (sludge volume index) was also detennined for
both 30 and 60 minutes o f settling and these results are shown in TabIe 4.5.
Several microplates using the Biolog technique were done during this experiment as
described in section 3.2.8. Three microplates using samples taken fiom the aerator on
three different days were incubated to detemine if the sludge population was changing on
a day-to-day basis. Three more microplates were done on samples taken from the aerator,
clarifier and a new shipment of sludge to compare the populations in each. The low-
resolution camera was used to take pictures of the microplates and the colour intensities
were analysed using PCA (principal cornponent analysis) as outlined in Materials and
Methods The scores for the first two principal components are displayed in Figure 4.8.
Figure 4.8: Scores for the First Two Principal Components for
Microplates Inoculated with Samples Taken trom Different Locations
and at Different Times for C T M P TME. (a = aerator, c = clarifier, s =
-
sludge, numbers represent different dates t(ll is the scores vector for
the first PC and t(2) i s the second).
The patterns for the three plates (and one replicate), prepared using samples taken from
the aerator over time, were visually very similar suggesting that the sludge population did
not Vary. In Figure 4.8a, the microplates done on two different days (denoted by B2 and
B6) are very similar while the plate done on another day (denoted by B3) is slightly
different. This could be due to differences in lighting or focus when the image was
obtained since at this tirne, a constant setup for taking the picture had not yet been
deveIoped. As for the plates taken fiom the aerator, clarifier and refrigerated sludge on
the same day (denoted by B6), the patterns for the aerator and clarifier were similar while
that for the sludge appeared different. This difference is demonstrated in Figure 4.8a (see
point B6s). This suggests that the new sludge requires an acclimation period.
In this chapter, the results and discussion for the treatment of eftluents fiom a Kraft mil1
in Quebec will be presented. Initially, a series of experiments were executed at various
hydraulic residence times (HRT's) in a CSTR with sludge recycle to determine the effect
of HRT on the reduction of chemical and biological oxygen demand (COD and BODI).
These results are given in section 5.1 and the expenments are compared in section 5.2.
Once these experiments were completed, the effect of a step input of black liquor at
various concentrations was deterrnined. These results are given in section 5 -3 and are
compared in section 5.4.
The reactor was operated at room temperature with aeration and agitation and the system
was allowed to mn until steady state was reached. Steady state was wnsidered to occur
when the concentration of total suspended solids and volatile suspended solids (TSS and
VSS) remained constant in the aerator for at least 3 HRT's. In al1 the experiments, the
biomass was the last parameter to reach steady state. The TSS and VSS in the aerator
followed the same trend in al1 experirnents and correlations were developed between
them. The same was tme for chemical oxygen demand and total organic carbon (COD
and TOC) and correlations were derived for these two parameters as well. The five-day
biological oxygen demand (BOD5) was determined in al1 but the first experiment (a 12 h
HRT) and correlations between COD and BOD5for the influent and effluent were found.
The average steady state values for the aerator and clarifier as well as the ratios of
VSSrrSS and COD/TOC are given in Table 5.1. The VSSlTSS was very similar between
the aerator and clarifier at about 0.84 and had a low standard deviation over a period of
time at about 0.0 1. This is a useh1 result since it is easier to measure TSS and the
VSS/TSS ratio c m be used to quickly estimate the VSS at any desired tirne and location.
Therefore, the measurement of VSS was limited only to the aerator in hture experiments.
The ratio of TOCKOD was also calculated and fiom Table 5.1 it is clear that they are
similar in the aerator and clarifier at about 0.35-0.36with a standard deviation of 0.05 in
the aerator and 0.03 in the clarifier. This ratio can be used to estimate the TOC if the
COD is known thus limiting the number of analyses that must be performed-
From Table 5.1 it can also be seen that the coefficient of variation (i.e. the ratio of the
mean to the standard deviation) for the TSS and VSS in the aerator was only about 10%
while it was much higher in the clarifier at about 50%. This can be explained by the fact
that t h e solids concentration in the clarifier fluctuated considerably during the mn. This
is a result of a number of factors that are discussed in section 5.2.
The pH, temperature, and DO generally remained constant throughout the expenment
(pH was about 7.0, T was 24OC. DO was 6.5 mg/l). The specific oxygen uptake rate
(SOUR), which is indicative of biomass activity, increased slightly after 80 h due to a
slight drop in the concentration of solids, but it generally remained in the range of around
10 mgO2/gVSS.h. The SVi was initially quite high as can be seen in Table 5.2. One
explanation for the p w r xttling at the beginning was due to inadequate wasting of the
sludge. Once wasting was improved, settling also improved. The significance of SVI is
discussed in greater detail later in this chapter in section 5.2.
Table 5.2: SVI for an HRT o f 12 h with Kraft Total Mill Effluent
Time (h) SVIM(mu@ S a (mUg)
47 343 343
93 3 73 362
Il4 332 3 19
136 3 18 315
Three microplates using the Biolog technique were inoculated with samples taken fkom
the aerator, clarifier, and refiigerated sludge. These were done during the steady state
period and the plates were incubated at rmm temperature for 48 h. The patterns for dl
three plates were visually very similar suggesting that the sludge population was constant.
PCA can only be performed on a data set of at least four or more observations; therefore,
it was not possible to mode1 these results within this experiment. However, it is possible
to compare these results to the next experiment using PCA and this is done in section 5.2.
l
c. Influent and A t riaor TOC I d Liiiucnt and Emuent BOD
250
Figure 5.2d represents the BOD5 for the system. Samples were taken fiom the influent
and t h e final effluent and both the BOD5 and COD were determined. The ratio, given in
Table 5.3, was about 0.2 for the influent and 0.03 for the effluent; they differed by a
factor o f slightly more than 10. These ratios are useful in estimating the BODs since the
BOD assay takes five days to produce results while COD is a relatively quick and easy
measure of substrate concentration. The average influent BODs was about 143 mg/l with
a standard deviation of about 23 which is reasonabiy low while the average effluent
BODs was about 7 mg/l with standard deviation of 1.4. This gives a reduction in BOD5
of 95% and the BOD uptake rate was 0.123 mgBOD/min-gVSS.
Table 5.3: Steady State Values and Ratios for an BRT of20 h with
Kraft Total Mill Emuent
1 Aemtor 1 Clarifiir 1 Inîluent €muent
Average Std Dev Average Std Dev Average Std
1
De\ 1
Average Std Do
The TSS and VSS clearly followed the same trend. Average steady state values for these
and other parameters are given in Table 5.3. The TSS and VSS in the aerator each had a
coefficient of variation of 16% while that for the clarifier was 3 1% which is lower than
the previous esperiment due to improvements in settling. Various ratios are also provided
in Table 5.3. Once again, the ratios were very simitar in both the aerator and clarifier.
The VSS/TSS was 0.82 with a standard deviation of only 0.01 for both the aerator and
clarifier while the TOCKOD was about 0.35 at al1 locations with a standard deviation of
about 0.03.
The trends of pH, temperature, DO, and SOUR were monitored for the system and these
generally remained constant (pH was 7.1, T was 24"C, DO was 6.8 mgA, SOUR was 8.4
mgg-h). The 30 and 60 minute SV1 tended to increase during the experiment as can be
seen in Table 5.4 (see Materials and Methods for the calculation of SVI). This is due, in
part, to the fact that SV1 is inversely proportional to TSS in the aerator. These results are
discussed in section 5.2.
Table 5.4: SV1 for an HRT of 20 h with Kraft Total Mill Emuent
For this experiment Biolog microplates were prepared on two different days. The low-
resolution camera was used to take pictures of the microplates. the colour intensities were
analysed using PCA, and a plot of the scores of the first two PC's i s given in Figure 5.3.
The first set o f plates (represented by the number 3 in the figure) were done at around
170 h, during the declining phase ofthe solids, using samples taken from the aerator (a).
clarifier (c). and refrigerated sludge (s) used as inoculum. The second set was done
during the steady state period at about 300 h (represented by the number 8) using samples
taken from the aerator and clarifier. For the first set, the patterns for each plate were
dityerent for each location: the plate inocutated with refiigerated sludge showed the
great est variation. The microplates taken during the neady state penod exhibited similar
patterns in both the aerator and clarifier. Comparing the patterns from day to day, the
microbial populations were slightly diflerent in the clarifier, but still very similar in the
aerator.
Figure 5.4d shows the influent and emuent BOD5. The BODs of the effluent was
constant throughout the experiment despite the fluctuation of the solids and COD
suçgest i ng t hat the system still achieved suficient BODs reduction even at non-steady
state conditions. Influent BODj was 227 mg/l and emuent BOD5 was 9 mgl, which
-eives a reduction of 96%. The ratios for BODICOD were 0.29 and 0.03 for the influent
and emuent respectively ( s h o w in Table 5.5). These values are analogous to the other
HRT experiments. The BOD uptake rate was 0.175 mgBOD1min-gVSS.
fI h hflwnt and AerPror COD
-
j j 0 CODa x CODi
I '
: w (b)
1
- - i
The ratio of VSS/TSS is about 0.79which is slightly lower than other HRT experiments.
Also, the TOCKOD ratio is about 0.46 for the aerator which is considerably higher than
the other HRT experiments. This could be due to the fact that a final steady state had not
been reached. The standard deviation is not presented here, a s there were not enough
points at the assumed steady state to accurately determine standard deviation.
The pH, temperature and DO generally remained constant throughout the expenment (pH
was 7.5, temperature was 24"C, DO was 6.5 mg/l). The SOUR increased significantly
(from 1 O to 25 mç02/gVSS-h)aAer 100 h due to the large drop in the concentration of
solids. SVI was deterrnined and these results are given in Table 5.6. The highest SV1
occurred du ring the period of lowest solids concentration.
Table 5.5: Stcady State Values and Ratios for an HRT of 8 h with
Kraft Total Mill Émuent
Clarifier
Average Average Average
Table 5.6: SV1 for an HRT o f 8 h with Kraft Total Mill Emuent
The sludçe age was adjusted to 10 days and the system was allowed to run for over 475 h
and reached steady state at about 350 h as s h o w in Figure 5.5a for the concentration of
solids (TSS and VSS). Figure 5.5 (b, c and d) shows that the COD, TOC and BODs
reached steady state within the first 50 h. The standard deviation for the COD in the
aerator was only 8.7 (shown in Table 5.7). At steady state the average COD for the
influent was 958 mgIl and for the aerator it was 276 mg/l corresponding to a reduction of
7 1%. The COD uptake rate was 0.3 11 mgCOD/min-gVSS. The steady state values for
BODS were about 199 mg/l for the influent and 2 mg/l for the effluent (shown in
Table 5.7). corresponding to a very high a reduction of 99%. The ratios for the influent
and emuent BODKOD were 0.21 and 0.008 respectively. The latter value is
considerabIy lower than ratios determined for other experiments suggesting that the
system was mnning at full efficiency in terms of BOD reduction. The BOD uptake rate
was 0.091 mgBOD/min.gVSS.
The TS S and VSS had very low standard deviations of only 1 13 and 78 respective1y
while the standard deviation for the TSS in the clarifier was also reasonably low
compared to other experiments at 1284 (see Table 5.7). The VSS/TSS was 0.8 with a
standard deviation of 0.02 and the TOCKOD was 0.34 with a standard deviation of 0.02.
These ratios are comparable to those obtained for the 12 and 20 h HRT experiments
Table 5.7: Steady State Values and Ratios for an HRT of 16 h using
Kraft Total Mill Emutnt
Acrator Clarifier Influent
A Std Dev Std Dcv Std Dcv
ab-
TSS cm@)
vss (ml$)
vssrrss
COD ( m g )
TOC <min
TOCICOD
BODs ( m u >
BODdCOD
VSSIACOD
-.c;
The averaee temperature was 2 2 T , pH was 6.7, DO was 6.5 mg/l, and SOUR was 9.6
mgOdgVSS-h. The SVI increased slightly during the experiment as the TSS decreased
for the same reasons as before. These results are shown in Table 5.8.
Table 5.8: SV1 for an HRT of 16 h with Kraft Total Mill Emuent
Time (h)
207 233 257 278 301 349 371 403 428 447 470
238 283 297 299 330 346 323 358 352 349'
268 274 288 282 287 313 332 306 339 330 327
Table 5.9 gives the values of the various ratios obtained in this research. The ratio of
VSSlTSS was uniform at about 0.80 showing that the biomass had the same carbon
content in al1 the experiments. The ratio for TOCKOD was relatively constant at about
0.34 for HRT's of 12, 16 and 20 h. This indicates that approximately 34% of the total
COD was organic material. A higher value of 0.46 was obtained for the 8 h HRT
experiment that did not reach a final steady state. Ratios for BODICOD were derived
separately for the influent and treated effluent as they typically differed by a factor of
almost ten. The BOD/COD ratio for the influent as received fiom the pulp miIl was fairly
constant averaging 0.22 (the ratio of the 8 h HRT was higher due to differences in
samples delivered to the lab on a weekly basis). The ratio of the COD to BOD of the
treat ed effluent varied between 0.03 to 0.0 1.
Table 5.9: Cornparison o f Ratios and Kinctic Parameters at Steady
State for Various HRT's
œ
ART (h)
4.
1
Sludge Age (days)
COD uptake (mg/min)lg
BOD uptaiic (mg/min)/g
% Reduction in COD
% Reduction in BOD
TOCICOD
BODICOD influent
BODICOD eMuent
VSSITSS
SOUR (mgOJgVSS-h)
S W O (mu@
S m a (ml/@
I
As a measure of sludge settling propenies, the sludge volume index (SVI) was
deterrnined at time intervals of 30 and 60 minutes. In most cases, the values did not
change significantly €rom 30 to 60 minutes of settling (for the equation for SVi see
Matenals and Methods). The steady state values for SV1 are given in Table 5.9.
Excluding the values for the 8 h experiment that failed to reach a final steady state, the
values for SV1 decrease linearly wit h an increase in HRT. This trend is s h o w in Figure
5.6. With long HRT's. nearly al1 the fed substrate is consumed by the cells resulting in
low BOD values. Hence, the cells exist in the endogenous growth phase characterised by
slow gronth rates that promote good settling. Furthermore, floc-forming bacteria are
only flocculant in a stationary growth phase such as the endogenous phase.
The SOUR is indicative of biomass activity. As the HRT increases the SOUR decreases
(see Figure 5.6)- because the biomass is receiving substrate at a reduced rate causing the
growth rate to slow down. With fewer celis being produced, less oxygen is utilised.
l X SOUR (rngOZ!gVSS-h) SI60 (ml'g) '
Comparing these scores to those of the variability study (section 3 -2.8C). it appears that
there were differences among microbial samples taken at each HRT on various days and
from various locations. The point that demonstrates the greatest difference is the point
for the sludge in the 20 h HRT experiment (point D3s in Figure 5.7). An explanation for
this is that in the 20 h KRT experiment, the sludge had been retngerated for longer than
usual. This is another indication that the sludge may have to go through a penod of
accli matisation. Comparing between expenments, there is a difference for each HRT
investigated; the points for each experiment are in distinct regions.
Figure 5.7: Scores for the Fint Two Principal Componcnts (t[l] and t[2]) for
Microplates Inoculated with Samplcs Taken from the Aerator, Clarifier and
Sludge at Various Times for HRT's of 12 and 20 h using Kraft Mill Effluent.
Within the 12 h HRT experiment, the scores for the microplates inoculated with sarnples
taken on the same day (denoted by C25) from the aerator, clarifier, and refrigerated
sludge indicate that the microbial population was the same in each location. For an HRT
of 20 h, the scores indicate that very slight differences exist between the aerator and
clarifier. This is likely due to the longer HRT that would favour the development of
slower growing organisms and give the sludge more tirne to shifl.
The reactor was kept ruming from the experiment with an HRT of 16 h so that the system
was already at steady state when the black liquor step was added. To initiate the step,
black liquor was added to the aerator, clarifier and feed stream simultaneously (see
Materials and Methods on the expenmental procedure). The initial TSS prior to the step
was about 1700 mg/l. Within 15 minutes of feeding the 5% solution the TSS increased to
1950 mg1 due to the greater concentration of solids in the black liquor (see Figure 5.8a).
The TSS in black liquor was about 600 mg/l causing a rise in the feed TSS fiom 20 mg/I
to 220 mgA.
After 16 h, the TSS dropped dramatically to 1100 mg/l indicating that the microbial
population had declined. However, after this initial dedine, the biomass population
began to grow at a rapid rate. In just over 60 h it increased to almost 4000 mg/l and
finally reached a steady state concentration of almost 5000 mg/l with standard deviation
of about 540. This trend can be seen in Figure 5.8a and the new steady state values are
shown in Table 5.10. From the Table it is obvious that at steady state the TSS and VSS
showed very little variation with coefficients of variation of less than 10%. The ratio for
V S S n S S was still in the same range as past experiments even in the presence of black
liquor. It also exhibited a low standard deviation.
! b hflucnt and AerPror COD
1
f
1
i
c. influent and Aerstor TOC i d hfluent and Emuent BOD
3000 -
8 ,
i
Just prior to the step, the COD concentration was 320 mg/l in the aerator and immediately
after the black liquor step the concentration increased to 5100 mg/l (the COD of the black
liquor was about 160 000 mg/]). This value gradually decreased to a steady state value of
about 3022 mg/]. Since the inlet COD was about 9085 mgIl this is a reduction of 67%,
which is comparable to past experiments without black liquor. A plot of COD is given in
~ a similar plot showing TOC for the aerator and influent and it
Figure 5.8b. Figure 5 . 8 is
follows the same pattern as the COD. The ratio TOCICOD (see Table 5.10) is in the
same range as the other experiments at 0.43 with a standard deviation of 0.03.The COD
uptake rate was quite high at 1.63 mgCOD/min-gVSS. This is much higher than any of
the previous experiments with just total miIl effluent suggesting that the activity o f the
sludge is greater. The BOD5 (Figure 5.8d) for the influent was 720 mg/l and for the
effluent at steady state it was 146 mg1 for a reduction of 80% (Table 5.10). The ratio of
BOD to COD is also given in Table 5.10. The BOD uptake rate was 0.145
rngBOD/min-gVSS.
Table 5.10: Steady State Values for a Stcp Input o f 5% Black Liquor
in Kraft Total Mill Effluent (HRT = 16 h)
-
Acrator
Std Dcv Std D& Std ih
5.8 1
235
0.01
34
0.01
The temperature, pH, DO and conductivity were monitored throughout the experiment.
The temperature remained between 22 and 23°C. AAer the black liquor step, the pH rose
from 7 . 3 to 8.6 and then stabilised at 8.2. Influent pH was adjusted to about 6 . 5 .
Conductivity was around 5.4-7.0 mS and the ratio of conductivity(S)/COD(mg/l) was 2.3
with a standard deviation of 0.2. Most significantly though, the DO dropped fiom 5.8 to
0.3 mg/]with the addition of black liquor. It was impossibie to increase the dissolved
oxygen concentration. This indicates that the cells were utilising the oxygen as fast as it
was being replenished and for this experiment, the oxygen became the Iimiting factor.
Because the DO was so low it was impossible to measure the SOUR. It was also
observed that the SVI decreased after the addition of black liquor, which suggests that the
biomass settled better. SV1 results are s h o w in Table 5.1 1. The SW30 just prior to the
addition of black liquor was about 350 ml/g (SWaowas 330 ml/g).
Table 5.1 1: SM for a Stcp Input of 5% Black Liquor in Kraft Total
Mill Efïluent (HRT = 16 h)
Timc (h) S M M (mi@ S m (mVg)
160 15 1 105
185 160 119
207 136 109
232 158 116
257 93 74
285 137 107
Bioiog microplates were done on samples taken just prior to, just afler, 14 h afler and
65 h after the step of 5% black liquor as well as at the new aeady state to see the response
of t h e population to the addition of black liquor. Quantitative results were obtained using
PCA as outlined in Materials and Methods and the scores are presented in Figure 5.9.
Figure 5.9: Scores for the First Two Principal Components for
Micropiates Taken at Various Times Before and After a Black Liquor
Step o f 5% in Kraft TME with an HRT of 16 h (t[l] is the Tint
principal component, t[21 is the second).
Prior to the step, the population pattern visually appeared normal (as compared to past
steady state patterns). Then, just afker the step, the pattern contained significantly fewer
wells showing colour development suggesting that many of the species had been reduced
to such an extent that they no longer demonstrated any affect on the colour development.
After only 14 h, the pattern became stronger (i-e.it contained more purple wells) and afier
65 h this new pattern was more pronounced. A final microplate was done when the
system reached steady state and this pattern differed fiom previous steady state patterns
without black liquor. These differences, demonstrated in Figure 5.9, suggest that there
was a shifi in the microbial population.
The steady state COD in the aerator pnor to the step was about 3 1 1 mg/l and afier the step
the COD increased to 3 150 mg/] (see Figure 5. lob). It then dropped gradually to a new
stead y state of 1644 mg/l in the aerator. The percent reduction o f the COD in aerator was
about 63% both before and afier the step. Figure 5 . 1 0 ~shows a similar trend in the TOC.
Ratios and steady state values are given in Table 5.12 for before and afier the black liquor
step. The ratio of TOC/COD compares reasonably well to other experiments. Also, the
ratio of conductivityKOD was 5.6 before the addition of black liquor and 2.7 after black
liciuor. This indicates that the presence of black iiquor causes a reduction in the
conductivity/COD ratio. The addition of black liquor also improved the COD uptake
rate, causing it to increase fiom 0.255 mgCOD/min-gVSS before the step. to 0.997
mgCOD/min-gVSS afier the step.
b M u e n t aod k r a o r COD 1
The influent and effluent BOD5 was determined and the results are s h o w in Figure
5.10d. Before the step, the steady state BOD5 was 207 mg/l for the influent and 6 mg/l
for the efl3uent as shown in Table 5. t 2. M e r the black liquor step, the BODS reached a
new steady state of 93 1 mgA for the influent and 58 mg/l for the effluent. The reduction
in BODs was initially 97% and dropped slightly to 94% after the step. The BOD uptake
rate was 0.095 mgBOD/min*gVSS before the step and 0.3 19 mgBOD/min*gVSS d e r the
step.
Table 5.12: Comparison o f Steady State Averages for a Black Liquor
Step of 2.5% in Kraft Total Mill Effiutnt (HRT = 16 h)
Acrator 1 Clarifier influent Effluent
TSS (mgm
vss ( m g )
VSSîïSS
COD (mgfi)
-roc( m m
TOCKOD
BODs (mgfi)
BOD$COD
Cond. (mS)
Cond./COD
(S/mgm
Table 5.13: SV1 for a Black Liquor Step o f 2.5% in Kraft Total Mill
Effluent (HRT = 16 h)
1 SV1 1 Time (h) 1
With the addition of black liquor, the temperature remained the same at 24°C while the
pH, conductivity, and SOUR increased and the DO decreased. The average values before
and afier t h e step respectively are as follows: pH increased from 6.8 to 8.3, conductivity
increased from 1.74 to 4.3 7 mS, DO decreased fiom 7.6 to 4.9 mg4 and SOUR increased
from 9.5 t o 31.1 mg02/gVSS-h. The notable increase in the SOUR after the step indicates
enhanced biomass activity. The SV1 decreased with the addition of black liquor (at 382h)
as in t h e previous experiment, confirming that the addition of black liquor does improve
setiling (see Table 5.13).
As in the previous experiment, Biolog microplates were done on samples taken just prior
to, just afler, 14 h afier and 48 h after the step with 2.5% black Iiquor as well as at the
new steady state. This was demonstrated quantitatively using PCA and the results are
shown below in Figure 5.11 as well as in section 5.4, where they are cornpared to other
experimental results.
Figure 5.1 1: Scores for the First Two Principal Components for
Microplates Taken at Various Times Before and After a Black Liquor
Step of 2.5% in Kraft TME with an HRT of 16 h (t[l J is the first
principal component, t[2) i s the second).
Prior to the step, the population pattem appeared normal (as compared to past steady state
patterns). The plate done just after the step was sirnilar to the one prior, but contained
several wells that were lighter than corresponding wells in the previous plate. This
suggests that a concentration of 2.5% black liquor did not cause as severe a stress to the
microbial population as did the 5% step. After only 14 h, the pattem became darker;
however, the 48 h pattern contained some newly purple wells that had not exhibited
colour development before. Finally, a microplate done at the new steady state produced a
pattern that was more similar to that for the regular steady state (with no black liquor).
The difference between the pattern obtained for the 48 h plate and the others seems clear
in Figure 5.1 1, where the 48 h plate is set apart from the other plates. The results indicate
that the microbial population changed durine the experiment.
5.3.3 Step Input of a WOSolution o f Black Liquor in TME
The reactor was run without black liquor (initial TSS was 2800 mgA) untii steady state
was reached at which time a step of 1% black liquor was added. Steady state results for
before and after the addition of black liquor are given in Table 5.14. The TSS fluctuated
substantially during the initia1 phase of the experiment due to problems with bulking and
plugging in the return sludge line, as can be seen in Figure 5.12a. Once the TSS
stabilised to some degree the step was initiated. This caused a slight increase in the TSS,
which then proceeded to increase tùnher to about 3 100 mg/[within one day of the step
input. The TSS then dropped, due to a plug in the retum sludge line, before increasing
again to 2800 mgl. Then the TSS slowly decreased until a steady state value of about
1700 mg1 was reached. The steady state TSS value before the black liquor step was
about 1600 mg/l so it was only slightly higher with the 1% black liquor. Ratios and
steady state averages are given in Table 5.14.
The COD for the aerator, clarifier and influent are s h o w in Figure 5.12b. The steady
state COD in the aerator prior to the step was about 300 mgl. Immediately afier the
black liquor step, the COD in the aerator increased and then decreased within one HRT to
930 mg/].From there, it reached the steady state value of about 797 mg/] within two
HRT's. The percent reduction for the COD in the aerator was 65% both prior to and after
shows a
the step input. This compares very well to previous experiments. Figure 5 . 1 2 ~
similar trend for the TOC. The ratios for TOCKOD were 0.54 and 0.42 before and afier
the step respectively. The COD uptake rate increased tiom 0.442 to 1.O 14
mgCOD/mi n.gVSS.
b hiluent aadAer.tor COD
2500 , 7
The influent and emuent BOD5 was deterrnined and the results are shown in Figure
5.12d. At steady state before the step, the BOD5 was 207 mg/l for the influent and IO
mgA for the emuent. At the new steady state after the black liquor step, the BOD, was
425 mg/l for the influent and 12 mg/l for the effluent. The reduction in BOD was initially
95% and increased slightly to 97% in the presence of black Iiquor. The BOD uptake rate
increased fiom 0.161 to 0.300 mgBOD/min-gVSS.
Table 5.14: Cornparison of Steady State Averages for a Black Liquor
Step of 1% in Kraft Total Mill Eîïîutnt (HRT= 16 h)
-- -
Acrator Influent
The average values for the system trends before and afier the step respectively were as
foIlows: temperature was 22°C and 24"C, pH was 6.96 and 6.87, conductivity was 1.49
and 2.42 mS. DO was 8.11 and 6.94 mgA. and SOUR was 15.76 and 34.77 rng02/gVSS-h.
Once again the SOUR dernonstrated a significant increase with the addition of black
liquor. The SV1 decreased sfightly with the addition of btack liquor (at 330 h) as
previously. but not to as great an extent (see Table 5.15).
Bioloç microplates were prepared for samples taken just pnor to, just afier, 16 h afler and
46 h afler the step with 1% black liquor to see the response of the population to the
addition of black Iiquor. PCA was used to quanti@ the results and a plot of the scores of
the first two principal components is given in Figure 5.13. These results are compared to
those of previous experiments in section 5.4.
c.
Cu
I I
O .
1 1
-
W.. '!
= !\ 1 Oh aft ;
Figure 5.13: Scores for the First Two Principal Components for
Microplates Taken at Various Times Before and After a Black Liquor
Step of 1% in Kraft TME with an HRT of 16 h (t[l] is the first
principal compontnt, t[2] is the second).
Prior to the step, the population pattem was slightly different from previous steady state
patterns. There were a few wells demonstrating colour development that norrnally do not,
implying that the microbial population in this experiment was different from those in past
experiments. Just afier the step, the pattem was very similar to the one just prior to it
suggesting that a concentration of 1% black liquor did not cause much of a shock to the
microbial population. This similarity is cleariy demonstrated in Figure 5.13, where the
points I Oh bef and 1 Oh aA are very close. AAer only 16 h, the pattern became darker and
both it and the 46 h pattem contained sorne newly purple wells that had not exhibited
colour development before. Finally, a microplate done at the new steady state showed a
pattem very similar to that of the 46 h plate. Overall, there was a difference between
corresponding microwells in the plates taken just prior to the step and at the new steady
state. The results indicate that there was a shift in the microbial population.
5.4 CO-!'PAFUSONOF BLACK LIQUOR STEP EXPERIMENTS
The following is a discussion of the results obtained during the black liquor step
experiments using total mil1 effluent and black liquor from the Krafî mill. The purpose of
these experiments was to determine the effect of black liquor on the microbial population
and its ability to reduce the substrate concentration. In general, higher black liquor
concentrations resulted in higher COD uptake rates while the system still managed to
reduce the organic content of the wastewater to the same degree as in the HRT
experiments. The black liquor also induced a shifl in the microbial population.
In most cases of the black liquor step inputs, the system still achieved satisfactory
reductions in COD and BOD5 as showri in Table 5.16. Al1 experiments obtained at least
6347% reduction in COD and these values did not Vary significantly within an
experiment, before and after the addition of black liquor. Two of the experiments also
attained high reductions in BODsin the presence of black liquor at 94% and 97% and
these values did not differ significantly from those of the steady state before the addition
of black liquor. The only experiment that did not achieve adequate BODsreduction was
that with the step of 5% black liquor, which only reduced the BODs by about 80%. This
may indicate that the system cannot handle such a large increase in fed substrate
concentration. This rnay be discussed by inspecting the COD uptake rate.
With the addition of black liquor, the COD uptake rate increased significantly as shown
in Table 5.16. A step input of 5% biack liquor resulted in an uptake rate of 1.6
mgCOD/min.gVSS, which was 5.2 times greater than that for the same system before the
addition of black iiquor. For the step of 2S%, the uptake rate was 1.O
mgCOD/min-gVSS, which was 3-9times greater then prior to black liquor. Finally, the
1% step gave an uptake rate of 1.O mgCOD/min-gVSS,which was 2.3 times greater than
before black liquor was added. The initial uptake rate for the 2.5% step was higher than
usual at 0.44 mgCOD/min-gVSS compared to a value of 0.26 mgCOD/min*gVSSfor the
2.5% step. To reduce the effects of this disparity, a plot of the COD uptake rate factor
(COD uptake rate at steady aate for the step concentration of black liquor divided by the
steady state uptake rate prior to the addition of black liquor) for increasing black liquor
concentrations is given in Figure 5.14. The figure illustrates that the COD uptake rate
factor increases with an increase in black liquor concentration.
The BOD uptake rate also increased with the black liquor steps as s h o w in Table 5.16.
For the step of 5%, the rate increased 1.6 times (fiom 0.09 to 0.15 m@OD/min~gVSS)-
The 2.5% step caused the rate to increase 3.4 times (fiom 0.10 to 0.32
mgBOD/rnin-gVSS)and the 1% step caused the rate to 1.9 times increase (fiom O. 16 to
0.30mgBOD/min-gVSS). These values show a large degree of variation since the BODs
was measured on samples taken From the final effluent after settling, where, as mentioned
above, the substrate concentration was much higher. Thus, these results are not as reliable
as those for the COD. Therefore, the BOD uptake factor is not plotted in Figure 5.14.
Figure 5.14: Effect of Black Liquor on the COD Uptake Rate for
Kraft Mill Efïluent
The black liquor causes a large increase in the feed substrate concentration and results in
a situation where the substrate is in excess. Therefore, the system is no longer operating
in the endogenous phase, but in the logarithmic growth phase where the growth rate of the
microbial population increases in order to utilise the available substrate. The rate of
metabolism is only limited by microbial generation and the population's ability to process
the substrate. In the case of the 5% step, where the BODs reduction was only 80%, the
biomass may have reached its maximum capacity for subsîrate reduction. One further
explanation for the removal of only 80% of the BOD could be that since the DO dropped
to levels below 0.5 mg4 and became the limiting factor in BOD reduction, the biomass
may have been prevented from reducing the BOD to the extent that it nomally does.
Table 5.16 shows the various ratios obtained for different black liquor concentrations.
The ratios did not change with the addition of black liquor. This implies that these ratios
may be used to estimate certain parameters in the system regardless of present conditions
at the treatment plant (ive.system upsets). However, one ratio that is present in the table
that had not been used in the previous experiments is that of conductivity1COD in
siemens (S) per mg of COD per litre. As demonstrated in the conductivity experiment as
weil as in literature (Kemeny and Baneqee, 1997), conductivity is a g w d indicator of
black liquor content in the wastewater. The ratio for conductivity1COD tends to decrease
linearly with an increase in black liquor concentration, as illustrated in Figure 5.14. This
is only true for treated effluent, as the conductivityXOD ratio appeared mainly constant
in the conductivity test using untreated effiuent. Another thing to note is that the ratio of
conductivity to COD was lower for the system containing black liquor at about 2.62
S/(mg/l) than it was in the experiments with just TME where it was about 5.46 S/(mg/l).
The same problems with the clarifier as discussed in the section for the HRT experiments
were also encountered here. In terms of settling. the SV1 was lower in the presence of
black liquor suggesting that the black liquor improves the settling properties of the
sludge. A reason for this may be that by adding black liquor, we are essentially
increasing the food to mass ratio ( F M ) and high FM loading selects for faster growing,
non-filamentous cells. The SVI did not, however, Vary significantly with black liquor
concentration. Thus the improvement in settling may have been a result of the different
type of substrate present and not the concentration of substrate. The composition o f black
Iiquor may have been the factor that selected for non-bulking bacteria. Funhermore, in
the parallel discussion on settling properties as a function of HRT, it was noted that high
HRT7s lead to improvements in SV1 because the system is operating in the endogenous
phase with slow growth rates. Therefore, with the increase in substrate concentration, the
system wouId enter the logarithmic growth phase characteri sed by high growth rates that
should, logically, lead to poor settling. Since this was not the case, the composition of the
black liquor must be the factor that caused improvements in settling.
As the fed substrate concentration increased (with the addition of black liquor), the SOUR
increased, causing the DO to decrease. This is a result of rapid biomass production as it
utilises the excessive amounts of substrate fed into the reactor, thus metabolising oxygen
at a higher rate. Steady state values for SOUR are given in Table 5.16.
Replicate plates done during a variability study (mentioned in section 3.2.8) are also
shown in Figure 5.15. The points for the scores of these plates are clustered together
indicating that the plates were very similar to one another. Concomitantly, the steady
state patterns just before the addition of black liquor show some degree of scatter. Thus,
each experiment started with a different microbial population. This was unavoidable
since the sludge was collected and delivered from the miIl on a weekly basis and
conditions, such as effluent composition, Vary at the mill. In order to determine the extent
of this difference, it would be necessary to nin the PCA on a set of plates inoculated with
several completely different microbial populations fiom separate systems. This would
show whether these plates (in Figure 5.15) were actually quite similar in comparison to
completely different populations.
Figure 5.15: Scores for the First Two Principal Components (t[l] and
t(2)) for Images of Microplates Inoculatcd with Samples from the
Aerator Treating Kraft Mill Effluent at Steady State before a Black
Liquor Step (Replicates are also shown).
After each step input, the system reached a new steady state. Scores for the microplates
done at each new steady state are shown in Figure 5.16 as well scores for the replicate
microplates. Once again, the points for the replicates are clustered together while the
points for the microplates done at the new steady States are scattered, indicating that
different black liquor step inputs resulted in different rnicrobial populations at the new
steady state. Thus, substrate concentration affects the microbial population. As
mentioned above, the extent to which these plates differ must be determined through a
comparison with several completely different microbial populations.
Figure 5.16: Scores for the First Two Principal Components (t[lJ and
t(2j) for Images of Microplates Inoculated with Samples from the
Aerator Treating Kraft Mill Eflluent at Steady State after a Black
Liquor Step (Replicates are also shown).
For each black liquor step experiment, the replicate scores were compared to the
experi mental scores in the same manner as above and in each case, the points for the
scores for each microplate were considerably more scattered in the figures than the points
for the repIicate plates. Thus, the microbial population undenvent a shifl with each
concentration of black liquor.
6. CONCLUSIONS AND RECOMMENDATIONS
In the experiments using various effluents fiom the CTMP mill, it was found that the ultra
high yield, paper machine, acid plant and w w d washing effluents al1 exhibited adverse
effects on the microbial population. In al1 cases, the growth of the biomass decreased
with increasing concentrations of these effluents in the total miIl effluent as did the
specific COD reduction. Increasing concentrations of each effluent also caused a
decrease in the reduction of COD in a11 cases, but the experiment using paper machine
emuent where the activated sludge still achieved suficient reduction in COD.
For the Krafi miIl effluent, the activated sludge system sufficiently reduced the substrate
in al1 cases investigated, regardless of variations in hydraulic residence time (KRT)or
black liquor concentration. The system achieved reductions of 63-72% in the COD and
90-99% in the BODs. The only exception occurred with a 5% black liquor step input
where only 80% of the BODs was reduced. However, typical upsets at the miIl rarely, if
ever, lead to black liquor levels as high as 5% of the total miIl effluent. Therefore, it may
be concluded that the sludge can handle most black liquor shocks typically experienced at
the mill.
The hydraul ic residence time had no effect on the substrate uptake rate. The COD uptake
rançed from 0.3 1 to 0.38mgCOD/min-gVSS at steady state for each HRT studied.
However, longer HRT's led to decreases in the specific oxygen uptake rate (SOUR) and
sludge volume index (SVI). Thus longer HRT's reduced the activity of the biornass and
produced a sludge with inferior settling properties. The step input of black liquor, on the
other hand, caused a significant rise in the substrate uptake rate and biomass activity. The
values for the substrate uptake rate and SOUR were much higher in the experiments with
black l iquor than in the experiments with only total mil1 effluent (TME). Furthemore,
the COD uptake rate increased (fiom 1.O to 1.6 mgCODImin-gVSS)with respect to rising
black liquor concentrations as did the SOUR indicating an increase in biomass activity.
The SVI decreased with a step input of black liquor, but did not change for different black
l iquor concentrations.
The Biolog microplate technique has been s h o w to be a useful tool for demonstrating
general changes in the microbial population. Although it is somewhat labour intensive
and costly, it is a valuable technique in that it gives an idea about the changes in the
population that cannot othenvise be detennined. The results fiom the Biolog microplates
confirmed that the microbial population experiences a shifl during a systern upset such as
a black liquor spill. However, to detennine the extent to which the population actudly
changed, it is recommended that principal component analysis be used to compare the
microbial populations in these experiments to several completely different populations
fiom separate systerns. The results obtained fiom these expenments do show that step
inputs of black liquor caused the microbial population to Vary over time, resulting in
different steady state populations. However, even though the microbial population
varied, the system still achieved suficient substrate reduction as stated above. In
conclusion, even though the black liquor induced a change on the microbial population,
the biomass proved to be very robust and capable of effectively treating the pulp mil1
effluent under many non-ideal conditions.
I 7. REFERENCES
Barr T. A., Taylor J. M., Duff S. J. B. (1996) Effect of HRT,SRT and Temperature
on the Performance of Activated Sludge Reactors Treating Bleached Kraft MiIl
Effiuent. Wat. Res. 30(4), 799-8 10
Benefield L., Lawrence D., Randall C. (1979) The Effect of Sludge Viability on
Biokinetic Coefficient Evaluation. J. WPCF 5 1(1), 1 87- 194
Betr Laboraties Inc. (1980) Betz Handbook of Industrial Water Conditioning. 8~ Ed.
Trevose, PA.
Bryant C. W., Avenell J. J., Barkley W. A., Thut R. N. (1992) The Removal of
Chlorinated Organics corn Conventional Pulp and Paper Wastewater Treatment
Systems. Wat. Sci. Tech. 26(1-2), 4 17-425
Builock C. M., Bicho P. A., Saddler J. N. (1996) The Influence of the High and Low
MoIecular Weight Fractions of a Bleach Kraft Mill Effluent on the Microbial
Activity of Activated Sludge. Wat. Res. 30(5), 1095-1 102
Chiu, S. Y., Erickson, L. E., Fan, L. T., Kao, 1. C. (1972) Kinetic Mode1
Identification In Mixed Populations Using Continuous Culture Data. Biotech.
Bioeng. 14, 207-23 1
10. Chiu, S. Y., Fan, L. T., Kao, 1. C., Erickson, L. E. (1972) Kinetic Behaviour Of
Mixed Populations Of Activated Sludge. Biotech. Bioeng. 14, 179-199
1 1 . Curds, C. R. (1973) A Theoretical Study Of Factors Influencing The Microbial
Population Dynamics Of The Activated-Sludge Process - 1 Effects Of Diurnal
Variations Of Sewage And Carnivorous Ciliated Protozoa. Wat. Res. 7, 1269-1284
3 3 . Mobius (1991) Nitrogen and Phosphorus Limits for Nutrient Deficient Industnal
Wastewaters. Wat. Sci. Tech. 24(3-4), 259-267
34. Morgan C. A., Wyndham R. C. (1 996) Isolation and Characterisation of Resin Acid
Degrading Bacteria Found in Effluent From a Bleached Kraft Pulp Mill. Can. J.
Microbiol. 42, 423-430
44. Templeton L. L., Grady C. P. L. Jr. (1988) Effect of Culture History on the
Determination of Biodegradation Kinetics by Batch and Fed-batch Techniques.
Journal WPCF 60(5), 65 1-658
Allen D. G., Liss S. N.(1996) Phenotypic Fingerprinting
45. Vittorio L., Gitbride K. A.,
of hlicrobial Communities in Wastewater Treatment Systems. Wat . Res. 30(5), 107%
1O86
APPENDIX A. SAMPLE CALCULATIONS I
Mill B EXP:
1OOO DATE:
100:3.0:1.0
TSS = -
(dn sample & filter weight (g) filter weight (g))* 1Oûû (mg)
(mglml) sample volume filtered (ml)
TSSo
PROPERTIES:
Phosplioric Acid:
Urea:
conc. N ( m u )=
conc. P (m@) =
Vurca = conc.N(m~)*60.06*Vt(ml) -
- 30*60.06*2000 = 4.21
(ml) 28*1000mVl*conc.of urea (mghi) 28*1000*30.514
APPENDIX B. PRINCIPAL COMPONENT ANALYSIS (PCA)
Data was collected using digital cameras and Visilog 5.1 image analysis software. The
software applied a macro to determine the average value in a 5x5 matrix of pixeis at the
centre of each well. This value represents the colour intensity of that well. Before PCA
could be perforrned, it was necessary to manipulate the data to improve the outcome of
the results. The method followed was one described by Garland and Mills, 1991.
Initially, the raw difference data was calculated by subtracting the value of the response
well fiorn that of the control well (C-R), reducing the control well to zero. Then, the data
was normalised by dividing by the average well colour development (AWCD) of that
plate: AWCD = x(C-R)/95. This step was necessary to account for differences in the
degree of colour development for each plate (Garland, 1996). This is referred to as the
transformed data. Once the transformed data is obtained for each plate the plates may be
compared to one another using PCA.
PCA is applied on a single s x tr data matrix X where s is the number of observations and
11 is the number of variables. These data are represented by a diffise and irregular swarm
of 11 points in s-space (Pielou, 1984). In this case, s is the number of microplates to be
compared and 11 is the number of microwells within a plate. Thus there are 95 variables.
The matrix X is set up by placing the intensity values for al1 the wells into a single row.
Therefore, rows represent individual plates (observations) while columns represent the
colour intensity in each microwell (variables). In this way, differences among the plates
may be determined. PCA cannot test for specific differences.
SIMCA-P statistical software follows the nonlinear iterative partial least squares
(NLPALS) method of calculating the principal components. This approach (as describe
by Geladi and Kowalski, 1986) consists of rewriting a data matnx X as a sum of linearly
independent matrices. These can in turn be expressed as produas of two vectors, a score
(column) vector f and a loading (row) vectorpT- Thus the PCA decomposition of the data
rnatrix X can be written as follows:
where a is the number of components or tpT products required to exactly describe the X
matris. In other words, each principal component (PC) is described by a score vector and
a loading vector. The first score vector is the one that passes through the data swarm in
the direction that describes the greatest portion of the variation among the data. The
second score vector is orthogonal to the first and passes through the data swarm in the
direction that describes the second greatest portion of the variation and so on so that the
first few components explain the greatest portion of the variation. A loading vector
shows how the variables are combined to forrn the scores (Le. it shows the responsible
variables). Altemately, X can be written as:
where T is a matrix whose columns are the score vectors /, while pT is a matrix whose
rows are the loading vectorspT. To illustrate the significance of the PCA decomposition,
consider a matrix X with two columns. This would correspond to a data matrix where
colour int~nsitiesfor only two corresponding wells are recorded.
Each column of the matrix X defined above contains colour intensity values for two
microwells in six different microplates. Each row of the matrix X consins of two entnes.
If one thinks of these as plane coordinates, then each row of the matrix X describes a
point in a plane whose axes XIand XIcorrespond to the colour intensities. Thus matrix X
consists of six points as illustrated in Figure B. Ib.
pi= cos 8,
pz= cos &
is nothing but a change of coordinate systems which allows one to describe the
information contained in the X matrix in a new coordinate system of reduced dimensions.
Once the direction cosines pl and p? of the principal component line are defined, only one
coordinate, t is needed to define each of the six points. This is illustrated in Figure B. la
and b.
The conclusions drawn fiom this two-dimensional example can be easily applied to cases
with more than two dimensions such as a fidl data set which contains 95 variables and as
many plates as desired. In addition, one can also easily see that the residual E is reduced
if X is approximated by more than one principal component (recall that a was defined
above as the number of components required to describe X without any error).
The NiPALS method calculates the principal components one at a tirne, while making
sure that each component is orthogonal to the previous one. The first component, given
by rPIT, is calculated fiom the Xmatrix so that:
after which the residual E2 is obtained and used to calculate the third component and so
on, until the a t h component for which E, = 0.
Step Intemretation
1. Let r, be any vector xj fiom X : t, = xj Any known vector can be used to stan off.
4. Recalculate t, : 1, = x ~ ~ . ~ , ~ Readjust
, scores in direct ion (X ont0 plv
5 . I f 1, ( 4 . )= f , (2.) stop, else goto 2. If f, is the same, the iteration has converged
and i p l T is the ith component; replace X by
its residual and start again at 1.
Rule 1: ~ ~ > l i m i t
where Q~ is the fraction of the total variation of X s that can be predicted by a
' = 1.O - PRESSISS
component: Q
Rule 2: at least fi has ~'v>lirnit
where K is the number of X-variables and Q ~ V
is the fraction of variation of a
variable, XI. : Q ~ V -
= (1.0 PRESS/SS)r for each variable k
Another criterion for significance is if the component's eigenvalue is greater than 2.0. It
should be noted here that SIMCA also computes the R
' or adjusted R~depending on
' was chosen and this is simply the fraction of
which is chosen. In this case the adjusted R
variance of al1 the Xs explained by the current component.
Once the vectors r and are found they can be ploned two dimensionally to reveal any
intrinsic patterns present in the data. Plots of the score vectors, i,. 12 and occasionally t;
are usuaily al1 that is required to demonstrate differences among plates. Score plots
display the observations in the new coordinate system defined by the PCA model where
each axis is mutually perpendicular. Loading plots ofp,, pl and if necessary, p3 should
also be included to fùlly describe the data. A plot of the loading vectors indicates which
of the variables are important in the approximation of the X matrix and corresponds to the
directions in the score plot.
This produces a matrix with an equal number of rows and columns. Then the eigenvalues
A, and eigenvectors 14, are computed for R. One can combine these n eigenvectors, which
are column vectors each with n elements, by letting them be columns in an 11 x n matrix
and obtain the orthogonal matrix of eigenvectors CI. At this point the eigenvectors may
be normalised such that their inner product or sum of squares equals 1;that is,
T
tt* M, =1
which is done by calculating a scaling factor k for each eigenvector by the following
equation:
The new coordinates can be computed by multiplying the orthogonal matrix O by the data
matrix X:
This results in a matrix o f the original s x rr dimensions only now each row s corresponds
to one principal component axis and the coordinates of each data point are no longer the
values used to describe colour development (Le. optical densities), but weighted sums of
al1 the values in the microwell (Ludwig and Reynolds, 1988, Pielou, 1984). Plotting the
first two principal components will give a two dimensional plot with n points (this is
equivalent to plotting the loading vectonpl and p? computed by the NIPALS method). If
al1 the points are plotted one would see that the pattern of the points relative to one
another is unchanged. Essentially, the only change that has occurred is that the entire
swarm as a single entity has been rotated about its centre of gravity, which is the origin of
the new coordinate fiame. In the same sense, plotting the eigenvectors u1 and u-, is
equivalent to plotting the scores rl and rz determined by the NIPALS method.